What is Together AI?
Serverless and dedicated inference for open models, with fine-tuning.
Who is Together AI best for?
Together AI is best for AI teams that want to run, fine-tune and host open models through one API.
Apps built on open models
One API serves more than 100 open models for chat, images, video, speech and transcription, billed by the token or by the item.
Teams leaving OpenAI's API
Existing OpenAI Python or TypeScript client code can call open models on Together without a rewrite.
Custom models on your own data
Fine-tune with LoRA or full training, then serve the result on a dedicated endpoint billed by the minute.
Training and heavy GPU jobs
GPU clusters rent H100, H200, B200 and B300 GPUs by the hour, with lower preemptible and reserved rates.
How much does Together AI cost?
Together AI is pay as you go on prepaid credits, with at least $5 to start: serverless models cost from $0.09 per million input tokens, fine-tuning from $0.34 per million tokens, GPU clusters from $1.99 per GPU hour and dedicated endpoints from $5.49 per GPU hour.
- Output from $0.14 to $15 per million tokens, by model
- Images, video, speech and transcription have their own rates
- Batch API up to 50% cheaper
- LoRA or full fine-tuning, supervised or DPO
- Full fine-tuning from $0.38 per million tokens
- Minimum charge of $4 to $60 per job, by model
- H100 at $1.99 preemptible or $3.99 on demand
- H200 $5.99, B200 $8.19, B300 $9.99 an hour on demand
- Reserved H100 from $3.19 an hour
- H100 at $5.49, B200 at $8.99 an hour on demand
- Billed by the minute for each running replica
- Other GPUs and reserved capacity through sales
Together AI is prepaid: there is no free trial, and the first credit purchase must be at least $5. Chat models run from $0.09 to $3.00 per million input tokens and $0.14 to $15.00 per million output tokens; Ternary Bonsai 27B is listed at $0.00. Provisioned throughput costs $0.05 per unit per minute, code sandboxes $0.0446 per vCPU hour plus $0.0149 per GiB of RAM an hour, and shared storage $0.16 per GiB a month. Reserved GPU cluster rates depend on the term, from $3.69 an hour for an H100 over 7 to 30 days down to $3.19 for 91 to 180 days. Full pricing on together.ai. We checked these prices on .
How do readers rate Together AI?
No one has rated Together AI yet. Be the first.
Sign in to rate. One rating per account.
What are the best Together AI alternatives?
Two tools on Recaply do Together AI's job. Both are open source, so you can run them on your own server.
Questions about Together AI
What is Together AI?
Serverless and dedicated inference for open models, with fine-tuning.
Who is Together AI for?
Together AI is best for AI teams that want to run, fine-tune and host open models through one API.
Is Together AI free?
No. Together AI has no free trial, and you need to buy at least $5 of credits before you can use the platform. One chat model, Ternary Bonsai 27B, is listed at $0, but calls to it still need an account with a positive credit balance.
How much does Together AI cost?
Together AI's paid plans are Serverless Inference at From $0.09 per million input tokens, Fine-Tuning at From $0.34 per million training tokens, GPU Clusters at From $1.99 per GPU per hour and Dedicated Inference at From $5.49 per GPU per hour.
Is there an open source alternative to Together AI?
Yes. Modular and Ollama are open source tools that do the same job, so you can run your own.
What are the best Together AI alternatives?
The closest alternatives on Recaply are Modular and Ollama.
Where can I try Together AI?
On its official website, together.ai.
Do Together AI credits expire?
No. Prepaid Together AI credits have no expiry date and cover API calls, dedicated endpoints and fine-tuning jobs. When the balance reaches zero, API access stops until you add more, and you can set up automatic top-ups with a card.
Tensorfuse
Modal
Upstash
Cloudflare
Neon
Prime Intellect