Fireworks AI pricing
4 plans read from fireworks.ai
Enterprise reserved instancePriced on request Contact salesReserved GPUs; multi-region deployments, custom optimizations, better performance, and BYOC compatibility Read 7 Oct 2026 · fireworks.ai
Managed trainingPrice not published Per 1M training tokens; rates vary by model size and training methodLoRA SFT, LoRA DPO, Full Param SFT, and Full Param DPO; rates range by model size Read 7 Oct 2026 · fireworks.ai
On-demand deploymentsPrice not published Per GPU second, with listed per-minute and per-hour ratesH100 and H200: $0.134/minute or $8.00/hour; B200: $0.217/minute or $13.00/hour; B300: $0.250/minute or $15.00/hour; GB300: $0.334/minute or $20.00/hour Read 7 Oct 2026 · fireworks.ai
Serverless inference — embeddingsPrice not published Usage-based, per input tokenPrices vary by embedding model parameter count Read 7 Oct 2026 · fireworks.ai
In short
- Fireworks AI has a free plan (Enterprise reserved instance).
- 1 plan is priced on request (Enterprise reserved instance).
What the alternatives costAll alternatives →
| Tool | From | Free plan | Free trial |
|---|---|---|---|
| Baseten | — | Yes | No |
| Cloudflare Stream | $1/mo | No | No |
| BentoML | — | Yes | Yes |
| fal | — | No | No |
| Runpod Serverless | — | Yes | No |
| Cohere North | — | No | No |
| Hugging Face Inference Endpoints | $0.06/mo | No | No |
| Modal | $250/mo | No | No |