Fireworks AI
Inference & Compute · Free + paid
Fast, affordable inference API for open-source models with sub-100ms latency
At a glance
- Made by
- Fireworks AI, Inc.
- Korean
- None (English and a few others only)
- API
- Yes
- Available on
- Web, CLI, API
What is Fireworks AI?
Fireworks is a training and inference platform that combines the latest open-source models with your own data to build specialized models. It supports everything from guided training to writing your own loss functions and RL loops, and deploys trained checkpoints to production in seconds. For inference, you can choose between serverless billed per token (OpenAI- and Anthropic-compatible), dedicated on-demand deployments, and reserved capacity. Fireworks Nexus is a drop-in alternative API that routes each task to the best open or closed model to reduce AI coding costs.
- Serverless inference: per-token pricing, no cold starts, OpenAI- and Anthropic-compatible
- On-demand dedicated deployment: per-second GPU billing, multi-region
- Reserved capacity
- Managed training: LoRA, full-parameter SFT, DPO
- Serverless Training API (shared trainer pool, LoRA)
- Self-written RL training loops (Training SDK)
- Fireworks Nexus: model routing cuts AI coding costs by 50–75%
- Model library (run open models with a single line of code)
Who uses Fireworks AI and for what
- Training and serving models for AI coding tools (Cursor case)
- Reducing latency with fine-tuning (Notion case: about 2 seconds to 350ms)
- Faster responses by hosting open-source models (Quora case: 3x)
Pricing
What you get for free — Unknown (Muse Glimmer 30B SFT training is free for a limited time until 10/31, no credit card required)
| Plan | Price | Main limits |
|---|---|---|
| Serverless Inference (prepaid credits) | Contact sales | Billed per input, output, and cached token; varies by model (e.g., GLM 5.3 Flash input $0.15/M, output $0.5/M) |
| Embeddings (up to 150M) | $0.008 | per 1 million input tokens |
| Managed Training LoRA SFT (up to 16B) | $0.5 | Per 1 million training tokens |
| On-Demand H100 80GB | $8 | Per GPU hour ($0.134 per minute) |
| On-Demand B200 180GB | $13 | Per GPU hour |
| On-Demand GB300 288GB | $20 | Per GPU hour |
| Enterprise | Contact sales | Contact sales |
Checked on the official pages on 2026-10-14. Prices and features can change, so check the official site again before paying. · Official pricing page
AI tools similar to Fireworks AI (alternatives)
Fireworks AI vs. alternatives at a glance
| AI | Pricing | Combos on the map | Connected tools | Free | Cheapest monthly | Korean |
|---|---|---|---|---|---|---|
| Fireworks AI | Free + paid | 0 | 5 | No | — | No |
| Akash Network | Free + paid | 0 | 17 | — | — | No |
| AWS Bedrock | Free + paid | 0 | 17 | No | — | Yes |
| AWS SageMaker | Free + paid | 0 | 17 | No | — | Yes |
| Azure ML | Free + paid | 0 | 17 | No | — | Yes |
0 of these 5 tools have a free plan.
Combo and connection counts are based on the relationships recorded on AI Galaxy Map (2026-10-26); they are not a quality ranking.
Fireworks AI on AI Galaxy Map
- One of 43 AI tools in the Inference & Compute field.
- Connected to 5 other tools — #16 in its field.
Frequently asked questions
Can I use Fireworks AI for free?
Unknown (Muse Glimmer 30B SFT training is free for a limited time until 10/31, no credit card required). Prices can change, so check the official site before paying.
What AI can I use instead of Fireworks AI?
Tools that do the same job, such as Akash Network, AWS Bedrock, AWS SageMaker, Azure ML, are connected to it on the map.
Can I use Fireworks AI in Korean?
None (English and a few others only)
Sources (official pages)
Checked on the official pages on 2026-10-14. Prices and features can change, so check the official site again before paying.