ProgrammableMarketing/together.ai (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 6
  • This quarter · 11
  • Big bets · 3
  • Wildcards · 3
  • How we can help
Generated Oct 22, 2026
  1. Programmable
  2. Playbooks
  3. Together AI
Together AI logo

growth playbook · Oct 22, 2026

together.ai (opens in a new tab)

Together AI growth playbook

Diagnosis

Together AI says it is growing 10x a year and serves over 400 trillion tokens a month. Its hiring shows how it plans to keep going: a first brand director, partner marketing for NVIDIA and cloud partners, ABM and field programs for AI-native and digital-native accounts, a technical marketing engineer for developer credibility, revenue systems that act on usage signals, and partnerships for day-zero model launches and distribution.

The gaps are on its own site: the homepage fails Core Web Vitals for real visitors while selling speed, customer proof is mostly startups although job posts name ElevenLabs, Salesforce and Zoom, and there is one solutions page and one partner page.

This playbook fixes speed, proof and usage signals first, then builds dinners, research nights, launch kits, migration guides, workload pages and partner routes, with a data report and a European conference as big bets.

Growth Score

8/10

Biggest Opportunity

Turn Together's research, usage data and customer results into proof for larger buyers, and act on the usage signals it already has.

Biggest Risk

Fireworks, Baseten, Groq and CoreWeave all claim speed and reliability, so without a distinct brand and enterprise proof, price competition decides deals.

01

Start here

The three moves to make first. Each one is written up in full below.

  1. 1

    Make the homepage as fast as the inference: real visitors wait 4.5 seconds and the page fails Core Web Vitals

    Export the Google Tag Manager container and list every tag with its owner and last use.

    High impact First result in 1 month
  2. 2

    Publish the enterprise stories your job posts already name: ElevenLabs, Salesforce and Zoom

    Email the three named customers' champions with a two-question request.

    High impact First result in 1 quarter
  3. 3

    Turn platform usage into a sales signal: alert the Sales Development Engineer when serverless spend jumps

    Pull last month's 50 fastest-growing serverless accounts.

    High impact First result in 1 month

02

Your idea portfolio

23 ideas, grouped by when to run them. Open any idea for how to run it: the steps, who to approach, copy you can use, and how to measure it.

  • 7 ideas are marked Unexpected: plays you are unlikely to have tried.

Every idea placed by the effort it takes against the impact it could have. Quick wins sit top left; select a number to open that idea. The filters below narrow the list by status, timing, category and what Programmable can build.

Effort vs impactEach number is an idea. Select one to open it.
Impact →
Quick wins · 5Big bets · 13Fill-ins · 4Reconsider · 1
Effort →

Showing 23 of 23 ideas

This week 6

Quick wins with a clear first action.

  • 01

    Together's homepage leads with "Faster inference 2x", yet Google's Chrome data shows real visitors wait 4.5 s for the main content and 236 ms for taps, failing both tests, and the site has slowed 39% in a year. The page is 5.2 MB, and 2.7 MB of that comes from 22 third-party services; Google Tag Manager alone blocks the browser for about a second. The new ABM & Campaign Manager becomes the business owner of the demand gen and ABM stack and will add more landing pages and tests, so this is the moment to set a budget before more tags go on.

    High impactMedium effort First result in 1 month Senior ABM & Campaign Manager

    First action

    Export the Google Tag Manager container and list every tag with its owner and last use.

    How to measure it

    Homepage LCP for real visitors

    Now
    –
    Target
    under 2.5 s
    By
    60 days

    Steps

    1. 1

      List tags

      Every tag, owner and purpose.

      GTM Engineer 2 days

    2. 2

      Cut and delay

      Remove unused tags; load the rest after interaction.

      GTM Engineer 1 week

    3. 3

      Trim JavaScript

      Drop unused scripts and heavy embeds from the homepage.

      Webflow developer 1 week

    4. 4

      Cache and lazy-load

      Long cache lifetimes; load lower sections on scroll.

      Webflow developer 3 days

    5. 5

      Set a budget

      Page weight and third-party limits for new pages.

      Senior ABM & Campaign Manager 1 day

    6. 6

      Re-test

      PageSpeed now, Chrome UX Report after 28 days.

      Webflow developer 1 month

    7. 7

      Measure

      Landing-page conversion before and after.

      Marketing Operations monthly

    First result
    1 month
    Cost
    Developer time only
    Owner
    Senior ABM & Campaign Manager

    Smallest test

    Remove or delay the five heaviest tags on the homepage and re-run PageSpeed.

    Why it works

    Audit every tag in Google Tag Manager with the GTM Engineer: keep one analytics tool, one ad pixel per network and the ABM tools that feed Salesforce, and remove or delay the rest until after the page is interactive. Ship the homepage with only the JavaScript it uses (936 KB is unused), set long cache lifetimes on Webflow assets, and load the model grid and research cards when they scroll into view. Set a performance budget (for example 1.5 MB and under 10 third parties on landing pages) that every new campaign tool must fit. Re-run PageSpeed and watch the Chrome UX Report for 28 days.

    Developers judge an infrastructure company by how its own site performs, and slow pages lose paid and ABM traffic before it converts. The fix mostly removes things, so it is cheap and lifts every campaign at once.

    Ready to use

    ChecklistHomepage speed checklist

    [ ] Export the Tag Manager container; list every tag with owner
    [ ] One analytics tool, one pixel per ad network
    [ ] Delay non-essential tags until after interaction
    [ ] Remove unused JavaScript (936 KB flagged)
    [ ] Long cache lifetimes on site assets
    [ ] Lazy-load model grid and research cards
    [ ] Budget: 1.5 MB, fewer than 10 third parties per landing page
    [ ] Re-run PageSpeed; check Chrome UX Report after 28 days

    Risks and how to handle them

    • Removing a tag breaks attribution

      Check each tag's reports with its owner before removal.

    • Webflow limits

      Move heavy widgets to lightweight embeds or load them on demand.

    Evidence

    • PageSpeed testGoogle PageSpeed for the together.ai homepage (October 21, 2026). Real Chrome visitors (Chrome UX Report, 75th percentile): main content appears in 4.5 s and taps respond in 236 ms, so the page fails Core Web Vitals on load and responsiveness (good is 2.5 s and 200 ms); site-wide, time to main content is 2.9 s and 39% slower than a year ago. Lab test on mobile: performance score 34, 5.2 MB and 193 requests, 2.7 MB of it from 22 third-party services; Google Tag Manager alone is 856 KB and blocks the main thread for about 1 s; 936 KB of JavaScript is unused.Source
    • Your websiteHomepage (October 21, 2026): "Build what's next on the AI Native Cloud"; headline claims "Faster inference 2x", "Lower cost 60%" and "Faster pre-training 90%"; banners for Together Post-Training, Together Link and on-demand B200s. The homepage model grid leads with Kimi K3, Qwen3.8 and GLM-5.3, while the Model Shaping menu still features Kimi K2.7 Code, GLM 5.1 FP4 and Llama 4 Maverick.Source
    • HiringSenior ABM & Campaign Manager (Demand Generation, San Francisco, posted September 8, 2026): owns ABM and integrated campaigns for "AI-native and digital-native companies running inference at scale"; runs "continuous testing and experimentation across channels — from ad creative and landing pages to targeting and channel mix"; is "the business owner of our demand gen and ABM stack" with RevOps; owns "account penetration and pipeline contribution against new logo goals".Source
    • HiringRevenue systems and sales: a GTM Engineer to build "signal-based workflows", "dynamic scoring models" and "centraliz[e] selling signals from multiple data sources" across Salesforce, HubSpot, Snowflake and Clay, and lead Salesforce CPQ; a Sales Development Engineer to "Convert users of the Together platform to broader commercial opportunities alongside Account Executives"; Strategic Account Executives in San Francisco, New York and London (EMEA inference).Source
  • 02

    Every job post says Together is "Trusted by Cursor, Decagon, ElevenLabs, Salesforce, and Zoom", but the 20 customer stories are almost all AI-native startups; only The Washington Post and Zomato are large established companies, and there is no story for ElevenLabs, Salesforce or Zoom. ABM now targets digital-native companies, whose buyers want proof from companies like them.

    High impactMedium effort First result in 1 quarter Head of Product and Solution…

    First action

    Email the three named customers' champions with a two-question request.

    How to measure it

    Enterprise stories live

    Now
    –
    Target
    5
    By
    2 quarters

    Steps

    1. 1

      Request

      Three named customers.

      Head of Product and Solution Marketing 1 week

    2. 2

      Interview

      One metric each.

      Head of Product and Solution Marketing 1 month

    3. 3

      Approve

      Customer legal review.

      Commercial Counsel 1 month

    4. 4

      Publish

      With an enterprise filter.

      Webflow developer 1 week

    5. 5

      Package

      One-page PDFs for sales.

      Head of Product and Solution Marketing 1 week

    6. 6

      Use in ABM

      Digital-native account plays.

      Senior ABM & Campaign Manager ongoing

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Head of Product and Solution Marketing

    Smallest test

    Get one quote with one metric from a named customer.

    Why it works

    Ask the named customers for a short story or a quote with one metric (latency, cost per token, time to launch). Add an "enterprise" filter to the customers page, and pair each story with the product it used. Give ABM and sales the stories as one-page PDFs.

    Recognisable enterprise names lower perceived risk for larger buyers, and the names are already being used in recruiting.

    Risks and how to handle them

    • Customers decline

      Offer an anonymised version or a quote only.

    Evidence

    • HiringTogether AI's description in its job posts: "10X growth year on year and has over a million developers"; "Trusted by Cursor, Decagon, ElevenLabs, Salesforce, and Zoom, Together serves over 400 trillion tokens a month." Open roles are concentrated in San Francisco (57) with teams in Amsterdam (9), London and Singapore.Source
    • Your websiteCustomer stories page: 20 stories, almost all AI-native companies (Cursor, Decagon, Cartesia, Hedra, Runware, Arcee AI, Deep Cogito, Yutori and others); the two large established enterprises are The Washington Post ("AI Independence with Reliable Inference") and Zomato (support bot that "doubled customer satisfaction"). Healthcare appears through Latent Health, XY.AI Labs and Slingshot AI. There is no story for ElevenLabs, Salesforce or Zoom, the names used in job posts.Source
    • HiringSenior ABM & Campaign Manager (Demand Generation, San Francisco, posted September 8, 2026): owns ABM and integrated campaigns for "AI-native and digital-native companies running inference at scale"; runs "continuous testing and experimentation across channels — from ad creative and landing pages to targeting and channel mix"; is "the business owner of our demand gen and ABM stack" with RevOps; owns "account penetration and pipeline contribution against new logo goals".Source
  • 03

    The Sales Development Engineer is hired to "Convert users of the Together platform to broader commercial opportunities", and the GTM Engineer to build "signal-based workflows". Pricing says most teams start on serverless and move to dedicated endpoints at scale. The moment a company's token spend climbs is the moment to offer provisioned throughput or dedicated inference.

    High impactMedium effort First result in 1 month GTM Engineer

    First action

    Pull last month's 50 fastest-growing serverless accounts.

    How to measure it

    Opportunities from usage alerts

    Now
    –
    Target
    40 a quarter
    By
    1 quarter

    Steps

    1. 1

      Join data

      Usage to Salesforce accounts.

      GTM Engineer 1 week

    2. 2

      Set thresholds

      Growth, spend and batch triggers.

      GTM Engineer 3 days

    3. 3

      Add savings estimate

      Dedicated vs serverless.

      Solutions Architect 1 week

    4. 4

      Alert

      Slack and Salesforce task.

      GTM Engineer 2 days

    5. 5

      Outreach

      Capacity review offer.

      Sales Development Engineer ongoing

    6. 6

      Review

      Conversion by trigger.

      Marketing Operations monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    GTM Engineer

    Smallest test

    Run the list by hand for two weeks.

    Why it works

    Join account usage to Salesforce by company domain. Alert the SDE when a company's serverless spend grows sharply week over week, crosses a threshold, or starts batch jobs, with the models used and an estimated saving on dedicated or provisioned throughput. The outreach offers a capacity review, not a pitch.

    Usage is the clearest buying signal in a usage-based business; the saving estimate gives a reason to talk.

    Risks and how to handle them

    • Developers dislike sales contact

      Offer engineering help and cost savings first.

    Evidence

    • HiringRevenue systems and sales: a GTM Engineer to build "signal-based workflows", "dynamic scoring models" and "centraliz[e] selling signals from multiple data sources" across Salesforce, HubSpot, Snowflake and Clay, and lead Salesforce CPQ; a Sales Development Engineer to "Convert users of the Together platform to broader commercial opportunities alongside Account Executives"; Strategic Account Executives in San Francisco, New York and London (EMEA inference).Source
    • Your websitePricing page: "Most teams start with serverless inference and move to dedicated endpoints at scale"; per-model prices per 1M tokens with a batch column, e.g. Kimi K3 $3.00 input and $15.00 output, GLM-5.3-Flash $0.15 and $0.50. The homepage says Batch Inference can "Scale to 30 billion tokens per model".Source
    • HiringTogether AI's description in its job posts: "10X growth year on year and has over a million developers"; "Trusted by Cursor, Decagon, ElevenLabs, Salesforce, and Zoom, Together serves over 400 trillion tokens a month." Open roles are concentrated in San Francisco (57) with teams in Amsterdam (9), London and Singapore.Source
  • 04

    The new partner marketer is hired for "NVIDIA and cloud partners", and the distribution lead for "cloud marketplaces". But the site has a single partner page (NVIDIA) and nothing on cloud partners or marketplaces, so buyers who prefer to buy through their cloud commitment cannot find the route.

    High impactLow effort First result in 1 month Senior Partner Marketing Mana…

    First action

    List every current partnership and marketplace route.

    How to measure it

    Pipeline through partner routes

    Now
    –
    Target
    15% of new pipeline
    By
    3 quarters

    Steps

    1. 1

      Inventory

      Current partnerships and routes.

      Senior Partner Marketing Manager 1 week

    2. 2

      Build the hub

      Webflow template.

      Webflow developer 1 week

    3. 3

      Write partner pages

      How to buy, billing, regions.

      Senior Partner Marketing Manager 2 weeks

    4. 4

      Add intake

      Become-a-partner form.

      GTM Engineer 2 days

    5. 5

      Launch rhythm

      Page and post per partnership.

      Senior Partner Marketing Manager ongoing

    6. 6

      Measure

      Partner-sourced pipeline.

      Marketing Operations monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Senior Partner Marketing Manager

    Smallest test

    Publish the hub with the NVIDIA page and a partner form.

    Why it works

    Create /partners with the NVIDIA page, a section for each cloud provider or marketplace Together is listed on or works with (how to buy, billing, regions), and a "become a partner" form for ISVs and data platforms. Every new partnership launch gets a page and a joint post on day one.

    Enterprises often buy through committed cloud spend; making the route visible removes a procurement objection.

    Risks and how to handle them

    • Partner approvals slow pages

      Agree page templates with partners up front.

    Evidence

    • HiringSenior Partner Marketing Manager (posted October 5, 2026): will "own and independently build the partner marketing function"; strategy "specially for NVIDIA and cloud partners"; runs "regular rhythms of business to drive co-marketing"; experience with "cloud hyperscalers (Azure, AWS, GCP) and/or NVIDIA" preferred.Source
    • HiringPartnerships: a Head of Distribution Partnerships to build "0-to-1 partnerships that drive token distribution" with "data platforms, cloud marketplaces, ISVs, and software ecosystems", including "marketplace listings, and case studies"; a Sr. Partnerships Manager, US Model Labs (posted October 6, 2026) to "Secure day-zero (or day-one) availability of the most anticipated model releases" and coordinate "launch marketing so we ship alongside the lab".Source
    • Your websiteNVIDIA partner page ("Two leaders. One platform."): "SOC II Type 2 certified, Zero Data Retention policy", "Dedicated clusters and private deployments for regulated industries", "OpenAI-compatible APIs to migrate in minutes". No page on the site covers a cloud provider partnership or a cloud marketplace listing.Source
    • Your websiteSitemap (654 www pages, read October 21, 2026): 282 model pages, 259 blog posts, 22 event pages, 20 customer stories, 18 model-provider pages and 6 GPU pages; one solutions page (Voice) and one partner page (NVIDIA); a "switch to open models" page, an open-model ROI calculator, a Startup Accelerator and a research credits program. The blog carries one-off head-to-heads such as "GLM-5.3 vs Claude Fable 5 on DeepSWE" and "DeepSeek V4 Pro 0813 vs GPT-5.6 Sol on DeepSWE".Source
  • 05

    Together has an ROI calculator that turns current closed-model spend into Together cost and 3-year savings, and a "switch to open models" page claiming up to 90% lower cost. The ABM manager will run gifting and meeting-maker programs for top accounts. A personal savings number is a stronger opener than a gift.

    High impactLow effort First result in 1 month Senior ABM & Campaign Manager

    First action

    Run the calculator for ten target accounts.

    How to measure it

    Meetings from ROI estimates

    Now
    –
    Target
    20% reply rate
    By
    1 quarter

    Steps

    1. 1

      Pick accounts

      Top 50.

      Senior ABM & Campaign Manager 2 days

    2. 2

      Estimate spend

      Public signals and sales notes.

      Sales Development Engineer 1 week

    3. 3

      Run calculator

      Per account.

      Senior ABM & Campaign Manager 3 days

    4. 4

      Write one-pager

      Model pairing and saving.

      Technical Marketing Engineer 1 week

    5. 5

      Send

      Email and LinkedIn.

      Senior ABM & Campaign Manager 1 week

    6. 6

      Validate

      Eval on their prompts.

      Solutions Architect ongoing

    First result
    1 month
    Cost
    Staff time only
    Owner
    Senior ABM & Campaign Manager

    Smallest test

    Send ten estimates this week.

    Why it works

    For the top 50 accounts, estimate closed-model spend from public signals and sales notes, run the calculator, and send a one-page estimate with the model pairing (for example a GLM or DeepSeek model against the closed model they use) and an offer to validate it on their own prompts.

    A specific number invites a reply, and the validation offer leads straight to a proof of concept.

    Risks and how to handle them

    • Estimates look presumptuous

      Present ranges and invite correction.

    Evidence

    • Your website"Switch to open models" page: "The top open models now match closed models on the benchmarks that matter, at up to 90% lower cost." The ROI calculator: "Enter your current spend and get your Together AI cost and your savings over 3 years."Source
    • HiringSenior ABM & Campaign Manager (Demand Generation, San Francisco, posted September 8, 2026): owns ABM and integrated campaigns for "AI-native and digital-native companies running inference at scale"; runs "continuous testing and experimentation across channels — from ad creative and landing pages to targeting and channel mix"; is "the business owner of our demand gen and ABM stack" with RevOps; owns "account penetration and pipeline contribution against new logo goals".Source
    • Your websiteSitemap (654 www pages, read October 21, 2026): 282 model pages, 259 blog posts, 22 event pages, 20 customer stories, 18 model-provider pages and 6 GPU pages; one solutions page (Voice) and one partner page (NVIDIA); a "switch to open models" page, an open-model ROI calculator, a Startup Accelerator and a research credits program. The blog carries one-off head-to-heads such as "GLM-5.3 vs Claude Fable 5 on DeepSWE" and "DeepSeek V4 Pro 0813 vs GPT-5.6 Sol on DeepSWE".Source
    • Your websitePricing page: "Most teams start with serverless inference and move to dedicated endpoints at scale"; per-model prices per 1M tokens with a batch column, e.g. Kimi K3 $3.00 input and $15.00 output, GLM-5.3-Flash $0.15 and $0.50. The homepage says Batch Inference can "Scale to 30 billion tokens per model".Source
  • 06

    The blog publishes separate posts such as "GLM-5.3 vs Claude Fable 5 on DeepSWE" and "DeepSeek V4 Pro 0813 vs GPT-5.6 Sol on DeepSWE". Each is useful once and then goes stale. The Technical Marketing Engineer owns "performance benchmarks" and wants developer credibility.

    High impactLow effort First result in 1 month Technical Marketing Engineer

    First action

    Collect the results from existing posts into one table.

    How to measure it

    Monthly leaderboard visits

    Now
    –
    Target
    50,000
    By
    2 quarters

    Steps

    1. 1

      Collect results

      From existing posts.

      Technical Marketing Engineer 3 days

    2. 2

      Design page

      Sortable table.

      Webflow developer 1 week

    3. 3

      Add cost

      Per task with Together prices.

      Technical Marketing Engineer 3 days

    4. 4

      Link

      From model pages and posts.

      Webflow developer 2 days

    5. 5

      Update

      Each model launch.

      Technical Marketing Engineer ongoing

    6. 6

      Promote

      X and Discord.

      Technical Marketing Engineer ongoing

    First result
    1 month
    Cost
    Staff time only
    Owner
    Technical Marketing Engineer

    Smallest test

    Publish the table as a blog post first.

    Why it works

    Publish a single page that ranks open and closed models on DeepSWE with cost per task, latency and the Together price, updated as models launch. Link every model page and comparison post to it, and post each update on X and Discord.

    A living leaderboard earns links and repeat visits, and AI assistants cite pages that stay current.

    Risks and how to handle them

    • Benchmark disputes

      Publish the method and raw results.

    Evidence

    • Your websiteSitemap (654 www pages, read October 21, 2026): 282 model pages, 259 blog posts, 22 event pages, 20 customer stories, 18 model-provider pages and 6 GPU pages; one solutions page (Voice) and one partner page (NVIDIA); a "switch to open models" page, an open-model ROI calculator, a Startup Accelerator and a research credits program. The blog carries one-off head-to-heads such as "GLM-5.3 vs Claude Fable 5 on DeepSWE" and "DeepSeek V4 Pro 0813 vs GPT-5.6 Sol on DeepSWE".Source
    • HiringTechnical Marketing Engineer (posted September 15, 2026): "Most of what makes Together AI so valuable for customers is invisible from the outside" (kernels, quantization, speculative decoding); a "creator role" producing "performance benchmarks, demo and explainer videos"; owns "technical communities on X, YouTube, Discord, Linkedin" and "amplifying the work of engineers and researchers who publish under their own names"; 20% in-person meetups.Source
    • Your websiteHomepage (October 21, 2026): "Build what's next on the AI Native Cloud"; headline claims "Faster inference 2x", "Lower cost 60%" and "Faster pre-training 90%"; banners for Together Post-Training, Together Link and on-demand B200s. The homepage model grid leads with Kimi K3, Qwen3.8 and GLM-5.3, while the Model Shaping menu still features Kimi K2.7 Code, GLM 5.1 FP4 and Llama 4 Maverick.Source

This quarter 11

The core plays for the next three months.

  • 07

    Together already runs Together Table dinners for AI leaders, AI natives and digital natives, and the new Field Marketing & Community Manager is hired for "repeatable account-based field programs" including dinners and roundtables, with international travel. Together has sales and engineering teams in London and Amsterdam.

    High impactMedium effort First result in 1 quarter Field Marketing & Community M…

    First action

    Book the next San Francisco and New York dates.

    How to measure it

    Opportunities from Together Table

    Now
    –
    Target
    6 per dinner
    By
    2 quarters

    Steps

    1. 1

      Set calendar

      Cities and dates.

      Field Marketing & Community Manager 1 week

    2. 2

      Pick accounts

      With sales.

      Senior ABM & Campaign Manager 1 week

    3. 3

      Book speakers

      One customer per dinner.

      Field Marketing & Community Manager ongoing

    4. 4

      Run

      Dinners.

      Field Marketing & Community Manager monthly

    5. 5

      Follow up

      Defined next step.

      Strategic Account Executive 1 week

    6. 6

      Report

      Pipeline and POCs.

      Marketing Operations monthly

    First result
    1 quarter
    Cost
    Dinner budget
    Owner
    Field Marketing & Community Manager

    Smallest test

    Run one London dinner.

    Why it works

    Run a monthly Together Table in San Francisco and New York and a quarterly one in London and Amsterdam, each for 12 to 16 leaders from target accounts, with one customer speaker and a research topic. Every dinner has a defined next step (a proof of concept or cost review) tracked in Salesforce.

    Small dinners create relationships that ads cannot, and a repeatable format scales with one manager.

    Risks and how to handle them

    • Same faces each time

      Rotate accounts and themes.

    Evidence

    • HiringField Marketing & Community Manager (San Francisco or NYC, posted October 5, 2026): "repeatable account-based field programs for priority accounts and segments, including dinners, roundtables, customer gatherings"; community "meetups, technical gatherings, workshops" built with "DevRel, Research, Product, Engineering"; measured on "account engagement, pipeline creation and acceleration, POCs"; 30-40% travel including international.Source
    • Your websiteEvents: Together Table dinners ("AI leaders NYC", "AI natives", "digital natives"), AI Native Conf 2026 in San Francisco (the CEO, Gamma's CEO and Cartesia's co-founder among speakers), NVIDIA GTC and GTC Berlin, AI Engineer Europe and World's Fair, ICML, a Kimi K3 webinar with Moonshot AI and a deep dive on NVIDIA Blackwell.Source
    • HiringTogether AI's description in its job posts: "10X growth year on year and has over a million developers"; "Trusted by Cursor, Decagon, ElevenLabs, Salesforce, and Zoom, Together serves over 400 trillion tokens a month." Open roles are concentrated in San Francisco (57) with teams in Amsterdam (9), London and Singapore.Source
    • HiringRevenue systems and sales: a GTM Engineer to build "signal-based workflows", "dynamic scoring models" and "centraliz[e] selling signals from multiple data sources" across Salesforce, HubSpot, Snowflake and Clay, and lead Salesforce CPQ; a Sales Development Engineer to "Convert users of the Together platform to broader commercial opportunities alongside Account Executives"; Strategic Account Executives in San Francisco, New York and London (EMEA inference).Source
  • 08

    Together's research (FlashAttention, ThunderKittens, ATLAS) is its strongest draw for developers, and the community role is asked to build meetups with Research and DevRel. The Technical Marketing Engineer is to translate research for developers in person 20% of the time.

    High impactLow effort First result in 1 quarter Field Marketing & Community M…

    First action

    Pick the first paper and author.

    How to measure it

    Attendees making API calls within 30 days

    Now
    –
    Target
    40%
    By
    2 quarters

    Steps

    1. 1

      Pick topics

      Six papers.

      Technical Marketing Engineer 1 week

    2. 2

      Book authors

      Research team.

      Field Marketing & Community Manager 2 weeks

    3. 3

      Venue

      Office or partner space.

      Field Marketing & Community Manager 2 weeks

    4. 4

      Run

      Monthly.

      Field Marketing & Community Manager monthly

    5. 5

      Credits

      Attendee codes.

      Developer Relations lead per event

    6. 6

      Record

      YouTube.

      Technical Marketing Engineer per event

    First result
    1 quarter
    Cost
    Venue and credits
    Owner
    Field Marketing & Community Manager

    Smallest test

    Run one research night in San Francisco.

    Why it works

    Host a monthly research night in San Francisco, alternating with New York and Amsterdam: one paper, one author, a live demo on Together, and a hands-on hour with credits. Record each talk for YouTube.

    Developers come for the people behind the research, and the events build preference that benchmarks alone do not.

    Risks and how to handle them

    • Researcher time

      Keep talks to 30 minutes.

    Evidence

    • HiringField Marketing & Community Manager (San Francisco or NYC, posted October 5, 2026): "repeatable account-based field programs for priority accounts and segments, including dinners, roundtables, customer gatherings"; community "meetups, technical gatherings, workshops" built with "DevRel, Research, Product, Engineering"; measured on "account engagement, pipeline creation and acceleration, POCs"; 30-40% travel including international.Source
    • HiringTechnical Marketing Engineer (posted September 15, 2026): "Most of what makes Together AI so valuable for customers is invisible from the outside" (kernels, quantization, speculative decoding); a "creator role" producing "performance benchmarks, demo and explainer videos"; owns "technical communities on X, YouTube, Discord, Linkedin" and "amplifying the work of engineers and researchers who publish under their own names"; 20% in-person meetups.Source
    • Your websiteHomepage (October 21, 2026): "Build what's next on the AI Native Cloud"; headline claims "Faster inference 2x", "Lower cost 60%" and "Faster pre-training 90%"; banners for Together Post-Training, Together Link and on-demand B200s. The homepage model grid leads with Kimi K3, Qwen3.8 and GLM-5.3, while the Model Shaping menu still features Kimi K2.7 Code, GLM 5.1 FP4 and Llama 4 Maverick.Source
    • Developer footprintGitHub togethercomputer: 111 public repositories, about 20,600 stars (OpenChatKit about 9,000) and 32 repos active in the last 90 days; the "together" Python SDK on PyPI is at version 2.42.0 (October 18, 2026); a Discord community and @togethercompute on X.Source
  • 09

    The Technical Marketing Engineer owns X, YouTube, Discord and LinkedIn and is asked to amplify "engineers and researchers who publish under their own names". Developers trust people more than company accounts.

    High impactLow effort First result in 1 quarter Technical Marketing Engineer

    First action

    List engineers and researchers who posted in the last month.

    How to measure it

    Engagement on developer content

    Now
    –
    Target
    +50%
    By
    2 quarters

    Steps

    1. 1

      List authors

      Opt-in group.

      Technical Marketing Engineer 1 week

    2. 2

      Write guide

      What to share.

      Technical Marketing Engineer 1 week

    3. 3

      Weekly loop

      Collect and amplify.

      Technical Marketing Engineer weekly

    4. 4

      Support

      Reviews and graphics.

      Design team weekly

    5. 5

      Track

      By author and topic.

      Technical Marketing Engineer monthly

    6. 6

      Recognise

      Internal shout-outs.

      Technical Marketing Engineer monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Technical Marketing Engineer

    Smallest test

    Amplify three engineer posts this week.

    Why it works

    Set up a weekly loop: collect what engineers publish, write a short company post that adds context, offer each author a review and graphics, and track engagement by author and topic. Give opt-in engineers a simple guide on what they can share.

    Personal posts reach further among developers, and support makes more engineers willing to publish.

    Risks and how to handle them

    • Disclosure mistakes

      Review posts touching unreleased work.

    Evidence

    • HiringTechnical Marketing Engineer (posted September 15, 2026): "Most of what makes Together AI so valuable for customers is invisible from the outside" (kernels, quantization, speculative decoding); a "creator role" producing "performance benchmarks, demo and explainer videos"; owns "technical communities on X, YouTube, Discord, Linkedin" and "amplifying the work of engineers and researchers who publish under their own names"; 20% in-person meetups.Source
    • Developer footprintGitHub togethercomputer: 111 public repositories, about 20,600 stars (OpenChatKit about 9,000) and 32 repos active in the last 90 days; the "together" Python SDK on PyPI is at version 2.42.0 (October 18, 2026); a Discord community and @togethercompute on X.Source
    • Your websiteHomepage (October 21, 2026): "Build what's next on the AI Native Cloud"; headline claims "Faster inference 2x", "Lower cost 60%" and "Faster pre-training 90%"; banners for Together Post-Training, Together Link and on-demand B200s. The homepage model grid leads with Kimi K3, Qwen3.8 and GLM-5.3, while the Model Shaping menu still features Kimi K2.7 Code, GLM 5.1 FP4 and Llama 4 Maverick.Source
  • 10

    The US Model Labs role is to secure day-zero availability of anticipated releases and coordinate launch marketing "so we ship alongside the lab". Together already runs launch webinars, such as Kimi K3 with Moonshot AI, and has rich model pages.

    High impactMedium effort First result in 1 quarter Sr. Partnerships Manager

    First action

    Write the kit as a checklist from the Kimi K3 launch.

    How to measure it

    First-week tokens on launched models

    Now
    –
    Target
    +30%
    By
    2 quarters

    Steps

    1. 1

      Write checklist

      From the last launch.

      Sr. Partnerships Manager, US Model Labs 1 week

    2. 2

      Templates

      Page, post, email.

      Technical Marketing Engineer 2 weeks

    3. 3

      Lab agreement

      Joint marketing terms.

      Sr. Partnerships Manager, US Model Labs per lab

    4. 4

      Run

      Each launch.

      Sr. Partnerships Manager, US Model Labs per launch

    5. 5

      Email users

      Similar-model users.

      Lifecycle marketing per launch

    6. 6

      Measure

      First-week tokens.

      Marketing Operations per launch

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Sr. Partnerships Manager, US Model Labs

    Smallest test

    Use the kit for the next launch.

    Why it works

    Standardise a launch kit: model page and pricing live at launch, a benchmark and cost post, a joint webinar with the lab within a week, a Discord and X thread, an email to developers using similar models and a sales note. Measure first-week tokens per launch.

    Launch weeks are when developers try new models; being ready on day zero wins the trial.

    Risks and how to handle them

    • Embargo leaks

      Restrict early access to named people.

    Evidence

    • HiringPartnerships: a Head of Distribution Partnerships to build "0-to-1 partnerships that drive token distribution" with "data platforms, cloud marketplaces, ISVs, and software ecosystems", including "marketplace listings, and case studies"; a Sr. Partnerships Manager, US Model Labs (posted October 6, 2026) to "Secure day-zero (or day-one) availability of the most anticipated model releases" and coordinate "launch marketing so we ship alongside the lab".Source
    • Your websiteEvents: Together Table dinners ("AI leaders NYC", "AI natives", "digital natives"), AI Native Conf 2026 in San Francisco (the CEO, Gamma's CEO and Cartesia's co-founder among speakers), NVIDIA GTC and GTC Berlin, AI Engineer Europe and World's Fair, ICML, a Kimi K3 webinar with Moonshot AI and a deep dive on NVIDIA Blackwell.Source
    • HiringTechnical Marketing Engineer (posted September 15, 2026): "Most of what makes Together AI so valuable for customers is invisible from the outside" (kernels, quantization, speculative decoding); a "creator role" producing "performance benchmarks, demo and explainer videos"; owns "technical communities on X, YouTube, Discord, Linkedin" and "amplifying the work of engineers and researchers who publish under their own names"; 20% in-person meetups.Source
    • Your websiteSitemap (654 www pages, read October 21, 2026): 282 model pages, 259 blog posts, 22 event pages, 20 customer stories, 18 model-provider pages and 6 GPU pages; one solutions page (Voice) and one partner page (NVIDIA); a "switch to open models" page, an open-model ROI calculator, a Startup Accelerator and a research credits program. The blog carries one-off head-to-heads such as "GLM-5.3 vs Claude Fable 5 on DeepSWE" and "DeepSeek V4 Pro 0813 vs GPT-5.6 Sol on DeepSWE".Source
  • 11

    The Model APIs PM is asked to close gaps that make it harder to "move workloads from proprietary model providers to Together", and the NVIDIA page promises "OpenAI-compatible APIs to migrate in minutes". Teams moving still face differences in tool calling, structured output and batching.

    High impactMedium effort First result in 1 quarter Senior Product Manager

    First action

    List the top five gaps customers hit when migrating.

    How to measure it

    Accounts migrated from proprietary APIs

    Now
    –
    Target
    100 a quarter
    By
    2 quarters

    Steps

    1. 1

      List gaps

      From support tickets.

      Senior Product Manager, Model APIs & Developer Experience 1 week

    2. 2

      Write guides

      Per SDK.

      Developer Relations lead 1 month

    3. 3

      Fix top gaps

      API changes.

      Senior Product Manager, Model APIs & Developer Experience 1 quarter

    4. 4

      Offer sessions

      Move in a day.

      Solutions Architect ongoing

    5. 5

      Promote

      Switch page and ABM.

      Senior ABM & Campaign Manager ongoing

    6. 6

      Measure

      Migrations and retention.

      Marketing Operations monthly

    First result
    1 quarter
    Cost
    Engineering time
    Owner
    Senior Product Manager, Model APIs & Developer Experience

    Smallest test

    Publish one guide.

    Why it works

    Publish a migration guide per major SDK and agent framework with the exact code changes, known behaviour differences and the closest open model by task. Offer a free "move in a day" session with a solutions engineer for teams above a spend level.

    Lowering switching effort turns the "switch to open models" message into actual migrations.

    Risks and how to handle them

    • Behaviour differences frustrate users

      Document them honestly.

    Evidence

    • HiringSenior Product Manager, Model APIs & Developer Experience (posted August 17, 2026): "Close the API and behavioral gaps that make it harder for customers to move workloads from proprietary model providers to Together"; owns ecosystem compatibility, voice and "Reimagine batch inference across the job lifecycle, developer experience, completion guarantees, pricing, and packaging".Source
    • Your websiteNVIDIA partner page ("Two leaders. One platform."): "SOC II Type 2 certified, Zero Data Retention policy", "Dedicated clusters and private deployments for regulated industries", "OpenAI-compatible APIs to migrate in minutes". No page on the site covers a cloud provider partnership or a cloud marketplace listing.Source
    • Your website"Switch to open models" page: "The top open models now match closed models on the benchmarks that matter, at up to 90% lower cost." The ROI calculator: "Enter your current spend and get your Together AI cost and your savings over 3 years."Source
    • Competitor researchRivals' own homepages: Fireworks, "Own Your Specialized Intelligence", training and inference that turn "open models into your specialized intelligence"; Baseten, "the fastest, most reliable inference platform"; Groq, "the premier neocloud for fast inference"; CoreWeave, "The Essential Cloud for AI".Source
  • 12

    Together's NVIDIA page says the two companies offer "the lowest cost-per-token on leading open models", Together attends NVIDIA GTC and GTC Berlin and ran a Blackwell deep dive. The partner marketer is hired to run "regular rhythms of business" with NVIDIA.

    High impactMedium effort First result in 2 quarters Senior Partner Marketing Mana…

    First action

    Draft the first quarterly plan.

    How to measure it

    NVIDIA co-marketing pipeline

    Now
    –
    Target
    $5M
    By
    3 quarters

    Steps

    1. 1

      Draft plan

      Quarterly.

      Senior Partner Marketing Manager 2 weeks

    2. 2

      Benchmark

      New hardware.

      Technical Marketing Engineer 1 month

    3. 3

      Webinar

      Joint.

      Senior Partner Marketing Manager quarterly

    4. 4

      GTC presence

      With a customer.

      Field Marketing & Community Manager per event

    5. 5

      Joint ABM

      Named accounts.

      Senior ABM & Campaign Manager quarterly

    6. 6

      Review

      Pipeline.

      Senior Partner Marketing Manager quarterly

    First result
    2 quarters
    Cost
    Co-marketing funds
    Owner
    Senior Partner Marketing Manager

    Smallest test

    Co-host one webinar.

    Why it works

    Agree a quarterly plan with NVIDIA: one joint benchmark on new hardware, one webinar, a GTC presence with a customer, and joint ABM for named accounts. Track joint pipeline.

    NVIDIA's reach and credibility amplify Together's performance claims.

    Risks and how to handle them

    • Partner priorities shift

      Hold a quarterly review.

    Evidence

    • HiringSenior Partner Marketing Manager (posted October 5, 2026): will "own and independently build the partner marketing function"; strategy "specially for NVIDIA and cloud partners"; runs "regular rhythms of business to drive co-marketing"; experience with "cloud hyperscalers (Azure, AWS, GCP) and/or NVIDIA" preferred.Source
    • Your websiteNVIDIA partner page ("Two leaders. One platform."): "SOC II Type 2 certified, Zero Data Retention policy", "Dedicated clusters and private deployments for regulated industries", "OpenAI-compatible APIs to migrate in minutes". No page on the site covers a cloud provider partnership or a cloud marketplace listing.Source
    • Your websiteEvents: Together Table dinners ("AI leaders NYC", "AI natives", "digital natives"), AI Native Conf 2026 in San Francisco (the CEO, Gamma's CEO and Cartesia's co-founder among speakers), NVIDIA GTC and GTC Berlin, AI Engineer Europe and World's Fair, ICML, a Kimi K3 webinar with Moonshot AI and a deep dive on NVIDIA Blackwell.Source
  • 13

    The Director of Brand is a new role, asked to build a narrative around "open, abundant AI" and "establish a brand tracker". Fireworks, Baseten, Groq and CoreWeave all claim speed or reliability, so Together's story must stand apart.

    High impactHigh effort First result in 2 quarters Director of Brand

    First action

    Pull 12 months of search volume for Together and four rivals.

    How to measure it

    Share of search among inference clouds

    Now
    –
    Target
    +5 points
    By
    3 quarters

    Steps

    1. 1

      Baseline

      Search, social, survey.

      Director of Brand 1 month

    2. 2

      Narrative

      With CEO and leaders.

      Director of Brand 1 month

    3. 3

      Creative

      Campaign.

      Director of Brand 1 month

    4. 4

      Launch

      Channels.

      Director of Brand 1 quarter

    5. 5

      Track

      Monthly.

      Director of Brand monthly

    6. 6

      Adjust

      Budget to winners.

      Director of Brand quarterly

    First result
    2 quarters
    Cost
    Brand budget
    Owner
    Director of Brand

    Smallest test

    Run the baseline survey with 300 developers.

    Why it works

    Before launching the campaign, baseline share of search, social mentions and a developer survey (awareness, consideration, associations) against the four rivals. Build the narrative around open models and research, launch, and report the tracker monthly.

    Measured brand work gets budget; without a baseline, the campaign cannot show results.

    Risks and how to handle them

    • Campaign looks generic

      Lead with research and open models.

    Evidence

    • HiringDirector of Brand (posted September 15, 2026), a new role: create "the brand narrative, grounded in the vision of the company to provide open, abundant AI and appealing to a technical audience"; build a brand campaign; "establish a brand tracker" for awareness and sentiment; own the brand budget.Source
    • Competitor researchRivals' own homepages: Fireworks, "Own Your Specialized Intelligence", training and inference that turn "open models into your specialized intelligence"; Baseten, "the fastest, most reliable inference platform"; Groq, "the premier neocloud for fast inference"; CoreWeave, "The Essential Cloud for AI".Source
    • Your websiteHomepage (October 21, 2026): "Build what's next on the AI Native Cloud"; headline claims "Faster inference 2x", "Lower cost 60%" and "Faster pre-training 90%"; banners for Together Post-Training, Together Link and on-demand B200s. The homepage model grid leads with Kimi K3, Qwen3.8 and GLM-5.3, while the Model Shaping menu still features Kimi K2.7 Code, GLM 5.1 FP4 and Llama 4 Maverick.Source
  • 14

    The Startup Accelerator gives "compute credits, engineering support, and GTM resources". The point of credits is to win companies that grow; the risk is losing them when credits end.

    Medium impactLow effort First result in 1 quarter Sales Development Engineer

    First action

    List accelerator companies above 50% credit use.

    How to measure it

    Accelerator companies converting to paid

    Now
    –
    Target
    50%
    By
    3 quarters

    Steps

    1. 1

      Track burn

      Per company.

      GTM Engineer 1 week

    2. 2

      75% trigger

      Cost review offer.

      Sales Development Engineer ongoing

    3. 3

      100% trigger

      Graduation offer.

      Sales Development Engineer ongoing

    4. 4

      Feature

      Stories and events.

      Head of Product and Solution Marketing quarterly

    5. 5

      Survey

      Why companies leave.

      Sales Development Engineer quarterly

    6. 6

      Review

      Conversion.

      Marketing Operations monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Sales Development Engineer

    Smallest test

    Call five companies near 75%.

    Why it works

    Track credit burn per company; at 75% used, the SDE offers a cost review and a committed plan; at 100%, a graduation offer. Feature graduating startups in stories and at AI Native Conf.

    A planned handoff keeps fast-growing startups, which are Together's core customers.

    Risks and how to handle them

    • Startups feel pressured

      Lead with help and savings.

    Evidence

    • Your websiteStartup Accelerator: "Get compute credits, engineering support, and GTM resources from Together AI's Startup Accelerator."Source
    • HiringRevenue systems and sales: a GTM Engineer to build "signal-based workflows", "dynamic scoring models" and "centraliz[e] selling signals from multiple data sources" across Salesforce, HubSpot, Snowflake and Clay, and lead Salesforce CPQ; a Sales Development Engineer to "Convert users of the Together platform to broader commercial opportunities alongside Account Executives"; Strategic Account Executives in San Francisco, New York and London (EMEA inference).Source
    • Your websiteCustomer stories page: 20 stories, almost all AI-native companies (Cursor, Decagon, Cartesia, Hedra, Runware, Arcee AI, Deep Cogito, Yutori and others); the two large established enterprises are The Washington Post ("AI Independence with Reliable Inference") and Zomato (support bot that "doubled customer satisfaction"). Healthcare appears through Latent Health, XY.AI Labs and Slingshot AI. There is no story for ElevenLabs, Salesforce or Zoom, the names used in job posts.Source
  • 15

    Voice is the only solutions page, yet the customer stories cluster around clear workloads: coding (Cursor), voice (Decagon, Cartesia), video and image generation (Hedra, Runware) and document parsing (XY.AI Labs). ABM targets companies "running inference at scale" by use case.

    High impactMedium effort First result in 2 quarters Head of Product and Solution…

    First action

    Draft the coding agents page from the Cursor story.

    How to measure it

    Contact-sales from workload pages

    Now
    –
    Target
    40 a month
    By
    2 quarters

    Steps

    1. 1

      Pick workloads

      Four.

      Head of Product and Solution Marketing 1 week

    2. 2

      Write pages

      Metric, models, deployment, cost.

      Head of Product and Solution Marketing 1 month

    3. 3

      Architectures

      Diagrams.

      Solutions Architect 1 month

    4. 4

      Publish

      Webflow.

      Webflow developer 1 week

    5. 5

      Use in ABM

      By workload.

      Senior ABM & Campaign Manager ongoing

    6. 6

      Measure

      Conversion.

      Marketing Operations monthly

    First result
    2 quarters
    Cost
    Staff time only
    Owner
    Head of Product and Solution Marketing

    Smallest test

    Publish the coding agents page.

    Why it works

    Build four workload pages with the best customer metric, the recommended models and deployment (serverless, provisioned or dedicated), a reference architecture and a cost example. Use them as ABM landing pages.

    Buyers search and evaluate by workload; pages that match their use case convert better than product pages.

    Risks and how to handle them

    • Overlap with product pages

      Link across, do not duplicate.

    Evidence

    • Your websiteSitemap (654 www pages, read October 21, 2026): 282 model pages, 259 blog posts, 22 event pages, 20 customer stories, 18 model-provider pages and 6 GPU pages; one solutions page (Voice) and one partner page (NVIDIA); a "switch to open models" page, an open-model ROI calculator, a Startup Accelerator and a research credits program. The blog carries one-off head-to-heads such as "GLM-5.3 vs Claude Fable 5 on DeepSWE" and "DeepSeek V4 Pro 0813 vs GPT-5.6 Sol on DeepSWE".Source
    • Your websiteCustomer stories page: 20 stories, almost all AI-native companies (Cursor, Decagon, Cartesia, Hedra, Runware, Arcee AI, Deep Cogito, Yutori and others); the two large established enterprises are The Washington Post ("AI Independence with Reliable Inference") and Zomato (support bot that "doubled customer satisfaction"). Healthcare appears through Latent Health, XY.AI Labs and Slingshot AI. There is no story for ElevenLabs, Salesforce or Zoom, the names used in job posts.Source
    • HiringSenior ABM & Campaign Manager (Demand Generation, San Francisco, posted September 8, 2026): owns ABM and integrated campaigns for "AI-native and digital-native companies running inference at scale"; runs "continuous testing and experimentation across channels — from ad creative and landing pages to targeting and channel mix"; is "the business owner of our demand gen and ABM stack" with RevOps; owns "account penetration and pipeline contribution against new logo goals".Source
    • Your websitePricing page: "Most teams start with serverless inference and move to dedicated endpoints at scale"; per-model prices per 1M tokens with a batch column, e.g. Kimi K3 $3.00 input and $15.00 output, GLM-5.3-Flash $0.15 and $0.50. The homepage says Batch Inference can "Scale to 30 billion tokens per model".Source
  • 16

    The Head of Distribution Partnerships is hired to drive "token distribution" through data platforms, cloud marketplaces and ISVs, with joint case studies. Data teams increasingly call models from inside their warehouse or app platform.

    High impactHigh effort First result in 3 quarters Head of Distribution Partners…

    First action

    Shortlist five platforms by fit.

    How to measure it

    Monthly tokens via distribution partners

    Now
    –
    Target
    5% of total
    By
    4 quarters

    Steps

    1. 1

      Shortlist

      Five platforms.

      Head of Distribution Partnerships 2 weeks

    2. 2

      Pitch

      Two partners.

      Head of Distribution Partnerships 1 quarter

    3. 3

      Integrate

      Engineering scope.

      Partner engineering 1 quarter

    4. 4

      Co-launch

      Joint story.

      Senior Partner Marketing Manager 1 month

    5. 5

      List

      Partner marketplaces.

      Head of Distribution Partnerships 1 month

    6. 6

      Measure

      Tokens served.

      Marketing Operations monthly

    First result
    3 quarters
    Cost
    Engineering time
    Owner
    Head of Distribution Partnerships

    Smallest test

    Pitch one platform.

    Why it works

    Pick two platforms where Together's open-model serving fills a gap, build the integration, co-launch with a joint customer story, and list Together in their marketplaces. Track tokens served through each partner.

    Distribution partners bring buyers Together would not reach directly, at low acquisition cost.

    Risks and how to handle them

    • Partner builds a competing service

      Secure commercial terms first.

    Evidence

    • HiringPartnerships: a Head of Distribution Partnerships to build "0-to-1 partnerships that drive token distribution" with "data platforms, cloud marketplaces, ISVs, and software ecosystems", including "marketplace listings, and case studies"; a Sr. Partnerships Manager, US Model Labs (posted October 6, 2026) to "Secure day-zero (or day-one) availability of the most anticipated model releases" and coordinate "launch marketing so we ship alongside the lab".Source
    • Your websiteNVIDIA partner page ("Two leaders. One platform."): "SOC II Type 2 certified, Zero Data Retention policy", "Dedicated clusters and private deployments for regulated industries", "OpenAI-compatible APIs to migrate in minutes". No page on the site covers a cloud provider partnership or a cloud marketplace listing.Source
    • HiringTogether AI's description in its job posts: "10X growth year on year and has over a million developers"; "Trusted by Cursor, Decagon, ElevenLabs, Salesforce, and Zoom, Together serves over 400 trillion tokens a month." Open roles are concentrated in San Francisco (57) with teams in Amsterdam (9), London and Singapore.Source
  • 17

    The PM is asked to "Reimagine batch inference" including "pricing, and packaging". Batch can "Scale to 30 billion tokens per model", and pricing lists a batch column, but there is no page that shows what batch is for.

    Medium impactMedium effort First result in 2 quarters Senior Product Manager

    First action

    Write the evaluation-run recipe.

    How to measure it

    Monthly batch tokens

    Now
    –
    Target
    +40%
    By
    2 quarters

    Steps

    1. 1

      Pick jobs

      Three recipes.

      Senior Product Manager, Model APIs & Developer Experience 1 week

    2. 2

      Write recipes

      Code and cost.

      Developer Relations lead 1 month

    3. 3

      Estimator

      Pricing page.

      Webflow developer 2 weeks

    4. 4

      Guarantee

      Completion terms.

      Senior Product Manager, Model APIs & Developer Experience 1 month

    5. 5

      Promote

      Email and docs.

      Technical Marketing Engineer ongoing

    6. 6

      Measure

      Batch usage.

      Marketing Operations monthly

    First result
    2 quarters
    Cost
    Staff time only
    Owner
    Senior Product Manager, Model APIs & Developer Experience

    Smallest test

    Publish one recipe.

    Why it works

    Create three batch recipes (evaluation runs, synthetic data generation, document backfills) with a cost example, completion guarantee and code. Add a batch cost estimator to the pricing page.

    Concrete jobs make batch an obvious choice and raise volume per account.

    Risks and how to handle them

    • Cannibalises real-time revenue

      Target new workloads.

    Evidence

    • HiringSenior Product Manager, Model APIs & Developer Experience (posted August 17, 2026): "Close the API and behavioral gaps that make it harder for customers to move workloads from proprietary model providers to Together"; owns ecosystem compatibility, voice and "Reimagine batch inference across the job lifecycle, developer experience, completion guarantees, pricing, and packaging".Source
    • Your websitePricing page: "Most teams start with serverless inference and move to dedicated endpoints at scale"; per-model prices per 1M tokens with a batch column, e.g. Kimi K3 $3.00 input and $15.00 output, GLM-5.3-Flash $0.15 and $0.50. The homepage says Batch Inference can "Scale to 30 billion tokens per model".Source
    • Your websiteHomepage (October 21, 2026): "Build what's next on the AI Native Cloud"; headline claims "Faster inference 2x", "Lower cost 60%" and "Faster pre-training 90%"; banners for Together Post-Training, Together Link and on-demand B200s. The homepage model grid leads with Kimi K3, Qwen3.8 and GLM-5.3, while the Model Shaping menu still features Kimi K2.7 Code, GLM 5.1 FP4 and Llama 4 Maverick.Source

Big bets 3

Larger plays that can change the trajectory.

  • 18

    Together serves "over 400 trillion tokens a month" to over a million developers. That gives it a unique view of which open models are used for what, how fast usage moves after launches, and how costs fall. No independent source has this view.

    High impactHigh effort First result in 3 quarters Director of Brand

    First action

    Agree the five charts with legal and data teams.

    How to measure it

    Report downloads

    Now
    –
    Target
    20,000
    By
    3 quarters

    Steps

    1. 1

      Pick questions

      Five charts.

      Director of Brand 2 weeks

    2. 2

      Privacy review

      Aggregation rules.

      Commercial Counsel 1 month

    3. 3

      Analyse

      Token data.

      Data team 1 month

    4. 4

      Write

      Report and explorer.

      Technical Marketing Engineer 1 month

    5. 5

      Launch

      Press and webinar.

      Director of Brand 1 month

    6. 6

      Repeat

      Twice a year.

      Director of Brand 6 months

    First result
    3 quarters
    Cost
    Analyst time
    Owner
    Director of Brand

    Smallest test

    Publish one chart on X.

    Why it works

    Publish an aggregated report twice a year: share of tokens by model family and task, adoption speed after launches, cost per task over time, and the move from closed to open. Launch with press, a webinar and a data explorer.

    Original data earns press and citations and makes Together the reference for open-model adoption.

    Risks and how to handle them

    • Customer data concerns

      Aggregate and suppress small groups.

    Evidence

    • HiringTogether AI's description in its job posts: "10X growth year on year and has over a million developers"; "Trusted by Cursor, Decagon, ElevenLabs, Salesforce, and Zoom, Together serves over 400 trillion tokens a month." Open roles are concentrated in San Francisco (57) with teams in Amsterdam (9), London and Singapore.Source
    • HiringDirector of Brand (posted September 15, 2026), a new role: create "the brand narrative, grounded in the vision of the company to provide open, abundant AI and appealing to a technical audience"; build a brand campaign; "establish a brand tracker" for awareness and sentiment; own the brand budget.Source
    • HiringTechnical Marketing Engineer (posted September 15, 2026): "Most of what makes Together AI so valuable for customers is invisible from the outside" (kernels, quantization, speculative decoding); a "creator role" producing "performance benchmarks, demo and explainer videos"; owns "technical communities on X, YouTube, Discord, Linkedin" and "amplifying the work of engineers and researchers who publish under their own names"; 20% in-person meetups.Source
    • Your websiteSitemap (654 www pages, read October 21, 2026): 282 model pages, 259 blog posts, 22 event pages, 20 customer stories, 18 model-provider pages and 6 GPU pages; one solutions page (Voice) and one partner page (NVIDIA); a "switch to open models" page, an open-model ROI calculator, a Startup Accelerator and a research credits program. The blog carries one-off head-to-heads such as "GLM-5.3 vs Claude Fable 5 on DeepSWE" and "DeepSeek V4 Pro 0813 vs GPT-5.6 Sol on DeepSWE".Source
  • 19

    AI Native Conf brought together founders and builders in San Francisco with the CEO and speakers from Gamma and Cartesia. Together has growing teams in Amsterdam and London and attends AI Engineer Europe and GTC Berlin.

    High impactHigh effort First result in 4 quarters Field Marketing & Community M…

    First action

    List 30 European AI-native companies to invite as speakers.

    How to measure it

    Conference-sourced pipeline

    Now
    –
    Target
    $10M
    By
    4 quarters

    Steps

    1. 1

      Pick city

      Amsterdam or London.

      Field Marketing & Community Manager 1 month

    2. 2

      Book speakers

      Customers and researchers.

      Field Marketing & Community Manager 1 quarter

    3. 3

      Sponsors

      NVIDIA and partners.

      Senior Partner Marketing Manager 1 quarter

    4. 4

      Invite

      ABM lists.

      Senior ABM & Campaign Manager 2 months

    5. 5

      Run and record

      All sessions.

      Field Marketing & Community Manager 1 day

    6. 6

      Follow up

      Sales.

      Strategic Account Executive 1 month

    First result
    4 quarters
    Cost
    Event budget
    Owner
    Field Marketing & Community Manager

    Smallest test

    Run a Together Table in Amsterdam first.

    Why it works

    Run AI Native Conf Europe in Amsterdam or London next year with European AI-native customers, and publish every talk from both events on YouTube within two weeks. Use attendance as an ABM tier.

    A flagship event makes Together the host of the AI-native community, not just a sponsor.

    Risks and how to handle them

    • Low European turnout

      Co-locate with an existing conference.

    Evidence

    • Your websiteEvents: Together Table dinners ("AI leaders NYC", "AI natives", "digital natives"), AI Native Conf 2026 in San Francisco (the CEO, Gamma's CEO and Cartesia's co-founder among speakers), NVIDIA GTC and GTC Berlin, AI Engineer Europe and World's Fair, ICML, a Kimi K3 webinar with Moonshot AI and a deep dive on NVIDIA Blackwell.Source
    • HiringField Marketing & Community Manager (San Francisco or NYC, posted October 5, 2026): "repeatable account-based field programs for priority accounts and segments, including dinners, roundtables, customer gatherings"; community "meetups, technical gatherings, workshops" built with "DevRel, Research, Product, Engineering"; measured on "account engagement, pipeline creation and acceleration, POCs"; 30-40% travel including international.Source
    • HiringTogether AI's description in its job posts: "10X growth year on year and has over a million developers"; "Trusted by Cursor, Decagon, ElevenLabs, Salesforce, and Zoom, Together serves over 400 trillion tokens a month." Open roles are concentrated in San Francisco (57) with teams in Amsterdam (9), London and Singapore.Source
    • HiringRevenue systems and sales: a GTM Engineer to build "signal-based workflows", "dynamic scoring models" and "centraliz[e] selling signals from multiple data sources" across Salesforce, HubSpot, Snowflake and Clay, and lead Salesforce CPQ; a Sales Development Engineer to "Convert users of the Together platform to broader commercial opportunities alongside Account Executives"; Strategic Account Executives in San Francisco, New York and London (EMEA inference).Source
  • 20

    Together promises "Dedicated clusters and private deployments for regulated industries" with SOC 2 Type 2 and zero data retention, and already has healthcare stories (Latent Health, XY.AI Labs, Slingshot AI). There is no page or program for regulated buyers.

    High impactHigh effort First result in 3 quarters Strategic Account Executive

    First action

    List the compliance questions from the last ten enterprise deals.

    How to measure it

    Healthcare and fintech pipeline

    Now
    –
    Target
    $8M
    By
    3 quarters

    Steps

    1. 1

      Gather questions

      From deals.

      Strategic Account Executive 2 weeks

    2. 2

      Security pack

      Documents.

      Security team 1 month

    3. 3

      Page

      Regulated industries.

      Head of Product and Solution Marketing 1 month

    4. 4

      Stories

      Healthcare focus.

      Head of Product and Solution Marketing 1 month

    5. 5

      ABM cluster

      Health-tech and fintech.

      Senior ABM & Campaign Manager 1 quarter

    6. 6

      Measure

      Pipeline.

      Marketing Operations monthly

    First result
    3 quarters
    Cost
    Staff time only
    Owner
    Strategic Account Executive

    Smallest test

    Publish the security pack.

    Why it works

    Package an offer for healthcare and financial services: private deployment, compliance documents, a security review pack and two healthcare stories. Build a page and an ABM cluster for health-tech and fintech accounts.

    Regulated buyers pay more and churn less once trust is established.

    Risks and how to handle them

    • Long sales cycles

      Start with health-tech startups.

    Evidence

    • Your websiteNVIDIA partner page ("Two leaders. One platform."): "SOC II Type 2 certified, Zero Data Retention policy", "Dedicated clusters and private deployments for regulated industries", "OpenAI-compatible APIs to migrate in minutes". No page on the site covers a cloud provider partnership or a cloud marketplace listing.Source
    • Your websiteCustomer stories page: 20 stories, almost all AI-native companies (Cursor, Decagon, Cartesia, Hedra, Runware, Arcee AI, Deep Cogito, Yutori and others); the two large established enterprises are The Washington Post ("AI Independence with Reliable Inference") and Zomato (support bot that "doubled customer satisfaction"). Healthcare appears through Latent Health, XY.AI Labs and Slingshot AI. There is no story for ElevenLabs, Salesforce or Zoom, the names used in job posts.Source
    • HiringSenior ABM & Campaign Manager (Demand Generation, San Francisco, posted September 8, 2026): owns ABM and integrated campaigns for "AI-native and digital-native companies running inference at scale"; runs "continuous testing and experimentation across channels — from ad creative and landing pages to targeting and channel mix"; is "the business owner of our demand gen and ABM stack" with RevOps; owns "account penetration and pipeline contribution against new logo goals".Source

Wildcards 3

Unexpected experiments, each with a cheap way to test it.

  • 21

    Together Link gives developers "a receipt showing what every session cost". Developers love to share savings and speed; a receipt comparing the session to a closed-model price is natural social content.

    Medium impactLow effort First result in 1 quarter Senior Product Manager

    First action

    Mock up the share card.

    How to measure it

    Link installs from shares

    Now
    –
    Target
    2,000
    By
    2 quarters

    Steps

    1. 1

      Design card

      Receipt image.

      Design team 1 week

    2. 2

      Build share

      In Link.

      Senior Product Manager, Model APIs & Developer Experience 2 weeks

    3. 3

      Add estimate

      Closed-model comparison.

      Senior Product Manager, Model APIs & Developer Experience 1 week

    4. 4

      Track

      Installs from links.

      Marketing Operations 1 week

    5. 5

      Feature

      On X.

      Technical Marketing Engineer weekly

    6. 6

      Review

      Shares and installs.

      Senior Product Manager, Model APIs & Developer Experience monthly

    First result
    1 quarter
    Cost
    Engineering time
    Owner
    Senior Product Manager, Model APIs & Developer Experience

    Smallest test

    Post three sample receipts on X.

    Why it works

    Add a "share" button that creates an image of the session receipt (cost, model, tokens and the estimated closed-model cost), with a link to install Link. Feature the best ones on X.

    User-made proof spreads further than ads and costs nothing per share.

    Risks and how to handle them

    • Estimates disputed

      Show the method.

    Evidence

    • Your websiteTogether Link page: "Run frontier open models in the harness you already use. One install command, automatic routing, and a receipt showing what every session cost."Source
    • Your website"Switch to open models" page: "The top open models now match closed models on the benchmarks that matter, at up to 90% lower cost." The ROI calculator: "Enter your current spend and get your Together AI cost and your savings over 3 years."Source
    • HiringTechnical Marketing Engineer (posted September 15, 2026): "Most of what makes Together AI so valuable for customers is invisible from the outside" (kernels, quantization, speculative decoding); a "creator role" producing "performance benchmarks, demo and explainer videos"; owns "technical communities on X, YouTube, Discord, Linkedin" and "amplifying the work of engineers and researchers who publish under their own names"; 20% in-person meetups.Source
  • 22

    The press page shows each headline twice with different dates (one story as both July 2026 and October 2025) and the press release list says "No items found". Journalists and analysts check it first.

    Medium impactLow effort First result in 1 week Director of Brand

    First action

    Fix the duplicate entries today.

    How to measure it

    Press page errors

    Now
    –
    Target
    0
    By
    2 weeks

    Steps

    1. 1

      Audit

      Entries and dates.

      Director of Brand 1 hour

    2. 2

      Fix

      Duplicates and dates.

      Webflow developer 1 hour

    3. 3

      Add releases

      Recent announcements.

      Communications 1 day

    4. 4

      Link badges

      Sources.

      Webflow developer 1 hour

    5. 5

      Assign owner

      Per launch.

      Director of Brand 1 day

    6. 6

      Review

      Monthly.

      Communications monthly

    First result
    1 week
    Cost
    $0
    Owner
    Director of Brand

    Smallest test

    Fix the duplicates today.

    Why it works

    Remove duplicates, fix dates, add the funding and launch announcements as press releases, and keep the AI 50 and InfraRed100 badges with links. Assign an owner to update it with each launch.

    A tidy press page supports the new brand role and costs an hour.

    Risks and how to handle them

    • Recurs

      Add to the launch checklist.

    Evidence

    • Your websitePress page: the "New & noteworthy" list shows each headline twice with different dates ("Open Source Growth Boosts Together AI, Hugging Face" as July 1 and June 23, 2026; the ATLAS speculator story as July 1, 2026 and October 10, 2025), the Press list below reads "No items found.", and "AI 50" and "InfraRed100" badges sit at the top.Source
    • HiringDirector of Brand (posted September 15, 2026), a new role: create "the brand narrative, grounded in the vision of the company to provide open, abundant AI and appealing to a technical audience"; build a brand campaign; "establish a brand tracker" for awareness and sentiment; own the brand budget.Source
  • 23

    The homepage leads with Kimi K3, Qwen3.8 and GLM-5.3, but the Model Shaping menu still features Kimi K2.7 Code, GLM 5.1 FP4 and Llama 4 Maverick. With 282 model pages and frequent launches, hand-edited menus fall behind.

    Medium impactLow effort First result in 2 weeks Webflow developer

    First action

    Update the menu by hand today.

    How to measure it

    Clicks from menus to newest models

    Now
    –
    Target
    +30%
    By
    1 quarter

    Steps

    1. 1

      Update today

      Manual fix.

      Webflow developer 1 hour

    2. 2

      Link collection

      Menu from catalogue.

      Webflow developer 1 week

    3. 3

      Sort rule

      Launch date or usage.

      Senior Product Manager, Model APIs & Developer Experience 1 day

    4. 4

      Test

      Next launch.

      Webflow developer per launch

    5. 5

      Add to kit

      Day-zero checklist.

      Sr. Partnerships Manager, US Model Labs 1 day

    6. 6

      Measure

      Clicks.

      Marketing Operations monthly

    First result
    2 weeks
    Cost
    Developer time
    Owner
    Webflow developer

    Smallest test

    Fix the menu by hand.

    Why it works

    Feed the menu's featured models from the same Webflow collection as the homepage grid, sorted by launch date or usage, so every launch updates navigation automatically.

    Current menus signal a fast-moving platform and send visitors to the models they came for.

    Risks and how to handle them

    • Menu changes too often

      Limit to six and pin key models.

    Evidence

    • Your websiteHomepage (October 21, 2026): "Build what's next on the AI Native Cloud"; headline claims "Faster inference 2x", "Lower cost 60%" and "Faster pre-training 90%"; banners for Together Post-Training, Together Link and on-demand B200s. The homepage model grid leads with Kimi K3, Qwen3.8 and GLM-5.3, while the Model Shaping menu still features Kimi K2.7 Code, GLM 5.1 FP4 and Llama 4 Maverick.Source
    • Your websiteSitemap (654 www pages, read October 21, 2026): 282 model pages, 259 blog posts, 22 event pages, 20 customer stories, 18 model-provider pages and 6 GPU pages; one solutions page (Voice) and one partner page (NVIDIA); a "switch to open models" page, an open-model ROI calculator, a Startup Accelerator and a research credits program. The blog carries one-off head-to-heads such as "GLM-5.3 vs Claude Fable 5 on DeepSWE" and "DeepSeek V4 Pro 0813 vs GPT-5.6 Sol on DeepSWE".Source
    • HiringPartnerships: a Head of Distribution Partnerships to build "0-to-1 partnerships that drive token distribution" with "data platforms, cloud marketplaces, ISVs, and software ecosystems", including "marketplace listings, and case studies"; a Sr. Partnerships Manager, US Model Labs (posted October 6, 2026) to "Secure day-zero (or day-one) availability of the most anticipated model releases" and coordinate "launch marketing so we ship alongside the lab".Source

03

How Programmable can help

Where an idea in this playbook is something Programmable builds or runs, it's listed here with what we'd do. Everything else is yours to run.

ProgrammableTalk

Fill the events in this plan

These ideas are events. Talk gets registrants for them and tells you who's coming and what their company does before the day.

Plan an event (opens programmable.talk in a new tab)
ProgrammableDigital

Show it, don't just describe it

These ideas need something to show: a demo, a video, an animation, graphics or a deck. Digital turns screenshots or the site into all of them, in the company's own brand.

Make a demo (opens programmable.digital in a new tab)
More example playbooks and reviews

More growth playbooks

Other companies we have researched, each read from what they are hiring for.

  • Nukleas growth strategyBack-office infrastructure for government contractors: a managed CUI enclave on Microsoft GCC High that prepares small contractors for…20 growth ideas
  • Haskell growth strategyAn integrated architecture, engineering, construction and consulting firm delivering food and beverage plants, healthcare, water and…21 growth ideas
  • Christina Savage, LLC growth strategyA Denver-area software consulting practice that solves hard automation problems: complex multi-step automations with AI steps, CRM and…23 growth ideas
  • airSlate growth strategyA portfolio of document and marketing software: SignNow (eSignature), pdfFiller and DocHub (PDF editing), altaFlow (workflow automation…23 growth ideas
  • Fasten Health growth strategyFasten Connect, an API and embeddable widget that lets patients share their own medical records with an app in minutes, from tens of…21 growth ideas
  • Pylon growth strategyAn agentic support platform for B2B companies: one inbox for Slack, Teams, email, chat, Discord, WhatsApp and phone, with AI agents that…21 growth ideas
  • Novae growth strategyManufactures open and enclosed trailers and truck bodies in aluminum and steel under 13 brands, including Sure-Trac, H&H Trailers, CAM…22 growth ideas
  • Prose growth strategyCustom, made-to-order haircare and skincare: an online consultation feeds Prose's Singular platform, which formulates each product and…21 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on October 22, 2026, and cites them throughout. Together AI did not commission, review or endorse it, and Programmable is not affiliated with Together AI. Company and product names are trademarks of their owners, used only to identify them. The ideas are suggestions and opinion, not professional, legal or financial advice, and not statements about how Together AI runs its business.

Something wrong, or do you represent Together AI? Email [email protected] and we will correct it or take it down.

30/60/90 roadmap

What to focus on in each month, and the ideas that carry it.

  1. Days 1 to 308 ideas

    Homepage speed, enterprise stories, usage signals, the partner hub, ROI outreach and a living DeepSWE leaderboard; fix the press page and model menus.

  2. Days 31 to 6012 ideas

    Together Table, research nights, engineer amplification, the day-zero launch kit, migration guides, NVIDIA co-marketing, the brand tracker, accelerator graduation, workload pages, distribution partnerships and batch recipes; shareable Link receipts.

  3. Days 61 to 903 ideas

    Commit to the big bets: the open-models report, AI Native Conf Europe and a regulated-industry offer.

The research behind the plan

3 research areas, each summarised in a few lines. Open any of them for the full detail.

Customers

What customers say

Customers value fast, low-latency inference. The most cited problem: closed-model costs at scale.

  • Lower cost than closed models
  • Latency for voice and agents
  • In their words: “open models”

Market

Competitive landscape

4 competitors compared: Fireworks, Baseten, Groq and CoreWeave.

  • Gap to exploit: Enterprise proof for open-model inference
  • Fireworks: Smaller compute footprint
  • Baseten: Less research brand

Hiring signals

Together AI is building its first full marketing and partner machine.

  • Senior ABM & Campaign Manager: Owns the demand gen stack and landing-page tests. Ideas 1, 5, 15.
  • Field Marketing & Community Manager: Dinners, roundtables and meetups. Ideas 7, 8, 19.
  • Senior Partner Marketing Manager: NVIDIA and cloud partners. Ideas 4, 12.

From the team that wrote this playbook

All products
  • ProgrammableAgencyWorking AI systems in days, not quarters (opens programmable.agency in a new tab)
  • ProgrammableInfoAn agent that finds, researches and asks firstNew accounts get 100,000 free enrichment credits with code BETA100K (opens programmable.info in a new tab)
  • ProgrammableDigitalScreenshots in, an animated product demo out (opens programmable.digital in a new tab)
  • ProgrammableWatchKnow the day a supplier falls out of compliance (opens programmable.watch in a new tab)
  • ProgrammableTalkGet registrants for your events, and know who’s coming (opens programmable.talk in a new tab)
  • ProgrammableMediaAn AI note taker, interpreter and webinar host for your calls (opens programmable.media in a new tab)
  • ProgrammablePageText what you need, your customer gets a page that works (opens programmable.page in a new tab)
  • ProgrammableCommunityForums with labelled AI members, never empty (opens programmable.community in a new tab)
ProgrammableMarketing
ExamplesLeaderboardsProductsTermsPrivacyCookiesDPALegalContact

© 2026 AI-Powered Growth Intelligence