Together AI
AI cloud for running, fine-tuning and training open models at scale
Overview
Together AI is a cloud platform for developers that serves open-weights language models such as DeepSeek, Qwen, GLM and Kimi through an OpenAI-compatible API, alongside image, video and speech models. It also sells batch inference, dedicated endpoints, fine-tuning and on-demand or reserved GPU clusters. Billing is prepaid per token or per GPU-hour, with a $15 minimum first purchase.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Serverless inference, per 1M tokens
- Serverless inference, per 1M tokens
- Serverless inference, per 1M tokens
- Serverless inference, per 1M tokens
- H200 from $5.99, B200 $8.19, B300 $9.99 per GPU-hour on demand
- GB200 and GB300 NVL72 on contact
- Single-tenant GPU instances
- Support for custom models
- Autoscaling and traffic-spike handling
Use cases
What you can produce with Together AI
- Open-weights models such as DeepSeek, Qwen, GLM and Kimi called through an OpenAI-compatible API
- Batch inference jobs at up to 50% lower cost than real-time calls
- A model fine-tuned on your own data (LoRA or full, SFT or DPO) and deployed to an endpoint
- Dedicated single-tenant GPU endpoints with autoscaling for production traffic
- On-demand, preemptible or reserved H100, H200, B200 and B300 GPU clusters for training
- Images, video, speech and transcription from hosted models under the same account and bill
Together AI is one of the most complete platforms for putting open models into production: a broad serverless catalogue behind an OpenAI-compatible API, published per-token prices, batch discounts, fine-tuning and GPU clusters from $3.99 an H100-hour, backed by serious inference research and an $800 million 2026 funding round.
It is strictly a builder's tool. There is no free tier or trial, billing is prepaid with a $15 minimum, and hosting is US-centric with no South-East Asian region listed, which regulated ASEAN buyers should weigh. Excellent for developers who want open-model inference and training under one bill.
What we found
Together AI, operated by Together Computer, Inc., is a San Francisco AI cloud set up in 2022 by Vipul Ved Prakash, Ce Zhang, Chris Ré, Tri Dao and Percy Liang. Tri Dao is the author of FlashAttention, and the company's research on inference kernels is a large part of its pitch. In July 2026 it announced an $800 million Series C; TechCrunch reported the round was led by Aramco Ventures at an $8.3 billion valuation, up from $3.3 billion at its Series B. The company also secured commitments for more than 500 MW of compute, and it names Cursor, Cognition, Decagon, ElevenLabs and Suno as customers.
The product is infrastructure for developers. Serverless inference serves open-weights language models such as DeepSeek, Qwen, GLM, Kimi, MiniMax, Gemma, Llama and gpt-oss through an OpenAI-compatible API. Next to them sit image, video, speech and transcription models, including closed ones such as Sora 2 and Veo, so earlier descriptions of an open-models-only catalogue are no longer accurate. Batch jobs, provisioned throughput, dedicated endpoints, fine-tuning, code sandboxes and storage fill out the stack.
Prices are published per million tokens and vary widely. On 14 October 2026 DeepSeek V4 Flash 0731 cost $0.14 in and $0.28 out, Llama 3.3 70B $1.04 each way, and Kimi K3 $3.00 in and $15.00 out. GPU clusters run from $3.99 per H100 per hour on demand, or $1.99 preemptible, to $9.99 for a B300. Batch work is listed at up to 50% less.
There is no free tier. Together's documentation says it offers no free trials, requires a $15 minimum first purchase and is fully prepaid, with API access suspended at a zero balance. One model listed at $0.00 does not change that. Hosting is US-centric, with no South-East Asian region on the pricing pages. For engineering teams moving open models into production it is one of the most complete options; non-developers will find nothing to use directly.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Together AI's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Together AI unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Together AI directly →
Spotted something out of date? Suggest an update →
Alternatives to Together AI
More in Agents & Automation
