Together AI
together.ai
Cloud platform for running and fine-tuning open-source AI models.
Quick Information
- Category:
- Coding
- Pricing:
- Pay-per-use
- Platforms:
- API
- Free Plan:
- No
- API Available:
- Yes
- Last Verified:
- September 2026
What Is Together AI?
Together AI is a cloud inference platform hosting over 200 open-source AI models from Meta (Llama), DeepSeek, Mistral, Alibaba (Qwen), Google (Gemma), and others — rather than creating its own models, it provides optimized infrastructure to run existing open-source models. Pricing is per-token and varies by model: small models under 4B parameters start around $0.10/million tokens, embedding models like M2-BERT run $0.008-$0.016/million tokens, and the broader catalog spans $0.05-$9.00/million tokens, with flagship models like DeepSeek V3.1 at $0.60/$1.70 per 1M tokens (input/output) and reasoning models like DeepSeek R1 at $3.00/$7.00.
Its dedicated deployment option — reserved GPU capacity priced per GPU-hour rather than per token, with the $4.19/hour H100 reserved rate dropping to $3.29/hour on a 91-180 day reservation — is genuinely one of the cheapest published managed-inference rates available for sustained training and large-batch workloads, making Together AI particularly cost-effective for teams with predictable, high-volume compute needs beyond simple pay-per-token serverless usage.
It's aimed at developers and teams wanting cost-effective access to open-source AI models without managing their own GPU infrastructure, with serverless pay-per-token pricing serving variable, development-stage workloads and reserved GPU-hour pricing serving sustained, high-volume production or training needs — the platform also offers a separate, application-only startup accelerator providing up to $50,000 in credits for qualifying companies. See our AI Infrastructure directory for related tools like Replicate and Fireworks AI.
What Can You Use Together AI For?
Open-source model inference
Runs open-source models from Meta, DeepSeek, Mistral, Alibaba, and Google via optimized cloud infrastructure.
Serverless pay-per-token deployment
Deploys models with no minimum commitment for variable, development-stage workloads.
Reserved GPU training and large-batch inference
Uses dedicated GPU-hour pricing for sustained, high-volume production or training workloads.
Together AI Features
- 200+ open-source models — Hosts optimized infrastructure for over 200 open-source AI models.
- Serverless pay-per-token — Provides pay-per-token pricing with no minimum commitment.
- Reserved GPU-hour pricing — Provides dedicated GPU capacity at competitive per-hour rates for sustained workloads.
Who Is Together AI For?
- Developers — Uses cost-effective access to a broad catalog of open-source models without managing GPU infrastructure.
- Teams with sustained compute needs — Uses reserved GPU-hour pricing for training and large-batch inference at competitive rates.
Together AI Pros & Limitations
What We Like
- Genuinely broad catalog — over 200 open-source models from major providers in one platform
- Competitive reserved GPU-hour rates ($3.29-$4.19/hour for H100) among the cheapest published managed-inference options
- Flexible pricing spans serverless pay-per-token to dedicated reserved capacity
Limitations
- Per-token pricing varies significantly by model, requiring careful cost estimation across the catalog
- Reserved GPU capacity requires longer-term commitment (91-180 days) to unlock the lowest hourly rates
- Doesn’t create its own proprietary models, relying entirely on the open-source ecosystem
Together AI Pricing
| Plan | Price | Main Features |
|---|---|---|
| Serverless (small models) | From ~$0.10/million tokens | Pay-per-token, no minimum commitment |
| Serverless (flagship models) | $0.60-$9.00/million tokens | Access to models like DeepSeek V3.1 and R1 |
| Reserved GPU (H100) | $3.29-$4.19/hour | Dedicated capacity for training and large-batch inference |
Pricing was last verified September 2026. Check the official website for the latest pricing.
AI Listing Jungle Ranking
- Global Rank:
- #213
- Overall Score:
- 67.8/100
- Monthly Users:
- 1M+
- Rank Change:
- ▲ 19
This ranking reflects AI Listing Jungle's own independent scoring methodology and is not an endorsement by Together AI or its company. How Our Rankings Work
Frequently Asked Questions
How is Together AI priced?
Per-token for serverless use, or per GPU-hour for reserved dedicated capacity.
What models does Together AI host?
Over 200 open-source models from Meta, DeepSeek, Mistral, Alibaba, Google, and others.
How much does reserved GPU capacity cost?
$4.19/hour for H100, dropping to $3.29/hour on a 91-180 day reservation.
Does Together AI create its own models?
No, it provides optimized infrastructure for existing open-source models.
Our Take
Together AI's reserved GPU-hour pricing ($3.29-$4.19/hour for H100) is genuinely among the cheapest published managed-inference rates for sustained training and large-batch workloads.
Its broad catalog of 200-plus open-source models provides real flexibility for developers wanting to avoid managing their own GPU infrastructure while still choosing from major model families.
Together AI Alternatives
Together AI Comparisons
Ready to try Together AI?
Visit the official website to get started.
Visit Official WebsiteAI Listing Jungle is an independent AI tools directory. Product names, logos, and trademarks belong to their respective owners.