Skip to content
Together AI
Together AI

Together AI for Startups

Together AI inference and fine-tuning credits, high-throughput hosted endpoints for Llama, Mistral, Qwen, DeepSeek, and dedicated GPU clusters at startup pricing.

AI PlatformsNo referral needed514 day reviewEasy to apply
Apply at Together AI
In a sentence

Together AI for Startups offers up to $1K–$25K in credits depending on partner to pre-seed, seed, series a startups, no VC referral required. Review takes 514 business days.

Max value
$1K–$25K in credits depending on partner
Credits valid
12 months
Regions
Global
Last verified
Apr 2026

About

Together AI runs the largest production fleet of open-weight model inference and offers dedicated GPU clusters (H100/H200) for training. The Startup Program offers credits for both serverless inference (per-token billing) and dedicated capacity, plus engineering support for fine-tuning and custom kernel optimization.

For founders building agents or RAG products on open models, Together's per-token cost on Llama 3.3 70B is among the lowest in the market, and the OpenAI-compatible API means existing OpenAI client code works with one base URL change.

Tiers

Startup tier
Inference credits + discounted dedicated capacity
  • Pre-Series B AI startup
  • Active build on open models
How to qualify: Apply via the startup program form or get a partner intro.

Eligibility

✓ You qualify if
  • AI startup
  • Pre-Series B
  • Open-model use case
✗ You don't qualify if
  • Series C+
  • Closed-model only (no need for Together)

How to apply

  1. 1
    Sign up
    Create a Together AI account.
  2. 2
    Apply via the startup form
    Submit company info and use case.
    www.together.ai/forms/startup-program
  3. 3
    Provision
    Credits are applied to your billing account.

What else you get

  • OpenAI-compatible API
  • 50+ open-weight models hosted
  • Fine-tuning service for popular models
  • Dedicated GPU clusters (H100/H200) at discount

What credits cover (and don't)

✓ Covered
  • Serverless inference
  • Fine-tuning
  • Dedicated GPU clusters
✗ Not covered
  • Closed-model APIs (use Anthropic/OpenAI)

Tactical tips

  • Tip 1.If you're already using the OpenAI SDK, switching to Together is a one-line base URL change for cost-savings on Llama-class models.
  • Tip 2.Together is a strong evaluation testbed before committing to dedicated capacity at scale.
  • Tip 3.For fine-tuning, Together's per-job pricing is often cheaper than rolling your own on Modal.

Common rejection reasons

  • Series C+
  • Production usage already at scale

Frequently asked about Together AI for Startups

Is Together AI for Startups free to apply?

Yes. Applying to Together AI for Startups does not cost anything and does not require giving up equity. Some programs require a payment method on file that activates only after credits are consumed or expire, check the program detail page for specifics.

How long does Together AI for Startups take to review applications?

Processing times are shown on the program detail page. Most programs reply within 1–3 weeks. Self-serve tiers (like AWS Activate Founders) can approve in 2–7 days; partner-referred tiers (like AWS Activate Portfolio) usually take 5–10 days.

Can I combine Together AI for Startups with other startup programs?

Most programs stack. The "Stacks well with" section on each detail page lists commonly combined programs. A few important exceptions: if you already claimed AWS credits via Brex or Mercury, your direct AWS Activate amount may be reduced.

What is the most common reason applications to Together AI get rejected?

The top rejection reasons are (1) using a personal Gmail/Outlook address instead of a company domain, (2) having a thin or placeholder website, and (3) mismatched information between the application and Crunchbase/Pitchbook. The tips section on the program page details program-specific factors.

More in AI Platforms

Related programs

See all →