Together AI
Together AI hosts open-source models with serverless, provisioned, and dedicated inference, plus fine-tuning and GPU clusters. Teams select models from a public catalog, run token‑priced serverless APIs or reserve throughput, and provision dedicated GPUs billed at documented hourly rates for production ML workloads.
The pricing pages list per‑model prices per 1M tokens (input/output) and per‑GPU hourly rates; fine‑tuning and sandbox compute costs are shown per token or per hour. Many individual model and GPU rates are displayed alongside batch and media pricing.Is Together AI right for you?
Good for
- Engineering teams that need production hosting and per-workload cost visibility for many open-source models and GPU options.
- Public catalog of many open‑source models and a model library for selection.
- Offers on‑demand serverless inference APIs to run models without managing infrastructure.
Keep in mind
- Displayed model prices may reflect lowest resolution/duration settings and can vary from final cost.
- Some enterprise hardware and reserved capacity items require contacting sales rather than showing public prices.
- Buyers seeking a single bundled monthly subscription price for all services; Together documents per-model token rates and per‑resource hourly charges instead of one flat plan.
Pricing and plan limits
The linked official pages list many precise prices (per 1M tokens, per GPU hour, per image/video/audio unit). The pages do not present a single all‑inclusive subscription price; some hardware and enterprise entries say "Contact sales."
More about capabilities and limits
Documented strengths
Public catalog of many open‑source models and a model library for selection.
Offers on‑demand serverless inference APIs to run models without managing infrastructure.
Detailed per‑model token prices and media rates shown, enabling per‑workload cost estimates.
Limitations to consider
Displayed model prices may reflect lowest resolution/duration settings and can vary from final cost.
Some enterprise hardware and reserved capacity items require contacting sales rather than showing public prices.
Choosing a plan
Use the published per‑model token and per‑GPU hourly rates to run a cost estimate for your target model, throughput mode (serverless vs provisioned), and expected token volume before deploying.
Based on the linked product documentation.
Available on
Web, Api.
Official sources
Alternatives to Together AI
Choose around the work you need to do.
Fireworks AI
Run, fine-tune, and serve models using both serverless per-token billing and on-demand dedicated GPU deployments to support development-through-enterprise training and inference workflows.
FriendliAI
Deploy open-weight and custom models for low-latency, high-throughput production inference with per-token Model API billing plus per-GPU dedicated endpoints and BYOG options for enterprise contracts.
