Fireworks
Inference and training platform for open models, fine-tunes, and production LLM serving.
- AI tool
- Freemium
- LLMs
- Coding agents
- ai agent frameworks
Pricing
Serverless pay-per-token with $1 free credits to start; on-demand GPU and reserved capacity; training priced per tokens or GPU hour.
Overview
Fireworks AI runs open and specialized models for inference and training. Serverless inference is OpenAI/Anthropic-compatible with Standard, Priority, and Fast tiers. On-demand and reserved deployments cover dedicated capacity and post-trained models. Training spans guided runs, config-led fine-tunes, and custom RL loops, with checkpoints deployable to production. Used for coding agents, composite models, and enterprise open-model serving.
Key features
- Serverless inference for popular open models
- On-demand and reserved GPU deployments
- Supervised, preference, and RL fine-tuning
- OpenAI-compatible API drop-in
- Multi-LoRA and post-trained model serving
- Embeddings and model catalog with published rates
Use cases
- LLM inference
- Model fine-tuning
- Open-source models
- Open-model API replacement
- Fine-tuned production models
Show all 7Show fewer
Alternatives
Last updated 2026-09-16