How much does Fireworks AI cost?
Usage price from
$0.01/minute
- Free plan
- Not stated
- Free trial
- Muse Glimmer 30B SFT free via UI and Serverless Training API until 11/30, no credit card…
- Usage price
- $0.01/minute
- Enterprise
- Contact sales
Read from fireworks.ai/pricing on Oct 29, 2026. Price changed? Tell usSee the line each price came from
In your favor
- Free trial. Muse Glimmer 30B SFT free via UI and Serverless Training API until 11/30, no credit card required
Watch out for
- Extra usage is billed. Usage above a Reserved Throughput reservation bills at standard Serverless rates
How you buy and get started
Self-serve + salesButtons on their site: Get started, Talk to experts, Contact us
- Sign up
- Self-serve signup "Start building in seconds, self-serve."
- Setup
- Zero setup for serverless "Get started in seconds with per token pricing, zero setup and no cold starts"
- When sales steps in
- Contact sales for enterprise deployments and region-restricted deployments "Region-restricted deployments are priced at a 1.5x premium."
What Fireworks AI does
Fireworks offers serverless per-token inference, managed and serverless training, and on-demand GPU deployments for open models. It also offers Nexus for routing between models, and enterprise controls such as zero data retention, SSO and regional endpoints.
- Good for
- Companies that want to run, route between and fine-tune open models, from startups to enterprises needing data residency and compliance controls.
- Channels
- Web dashboard, API, CLI (firectl)
- Main product
- Fireworks inference platform (serverless, on-demand and reserved inference, plus training)
- Also includes
- Serverless inference, Embeddings, Managed training (SFT and DPO), Serverless Training API, Dedicated Training API, On-demand GPU deployments
- Connects to
- Claude Code, ChatGPT, Codex, Cursor, Microsoft Foundry, Okta, Microsoft Entra ID, Google Workspace
- Public API
- Yes
- MCP server
- Not stated
- Public API access
New open models arrive on Fireworks on launch day, behind OpenAI- and Anthropic-compatible APIs.
- Long context window
1,048,576 Context
- Fine-tuning support
Customize open models with your own data with minimal setup
- Tool and function callingNot stated
- Image and audio input
Fine-tuning with images (VLM supervised fine-tuning) is also billed per 1M tokens.
- SOC 2 compliance
SOC 2 Type 2
- Zero data retention option
Fireworks keeps no prompts or generations for open models by default.
- VPC or on-prem deployment
Run Fireworks-managed inference on GPUs in your own cloud account or data center.
The same 8 checks appear on every tool in Large language models. Each tick quotes the vendor's own site.
Security and your data
| Question | Their answer | Where it says so |
|---|---|---|
| Trains AI on your data? | no | "We do not use your prompts, training data, or API inputs to train or improve our AI models without your explicit opt-in."privacy |
| HIPAA (signs a BAA) | available | "ZDR by default, SOC 2 Type II, HIPAA, and ISO 27001 certified."enterprise |
| SOC 2 | SOC 2 Type 2 | "SOC 2 Type 2"home |
| GDPR | GDPR listed in compliance | "GDPR"security |
| CCPA | Not stated | |
| PCI | Not stated | |
| ISO 27001 | ISO 27001 | "ISO 27001"home |
| Single sign-on | Custom SSO and SCIM (enterprise) | "Sign in through your SAML 2.0 or OIDC identity provider and enforce SSO for every user."enterprise |
| Trust center | https://trust.fireworks.ai/ | "Visit trust center"enterprise |
Security answers are the vendor's own claims, quoted. We don't audit them.
Every plan
| Plan | Monthly | Billed yearly | Included | Over the allowance | Where it says so |
|---|---|---|---|---|---|
| Embeddings up to 150M paramsper 1M input tokens | $0.01per minute | – | – | – | "$0.008" |
| Embeddings 150M-350M paramsper 1M input tokens | $0.02per minute | – | – | – | "$0.016" |
| Embeddings Qwen3 8Bper 1M input tokens | $0.10per minute | – | – | – | "$0.1" |
| On-demand H100 80 GBper GPU hour | $8per minute | – | – | – | "H100 80 GB GPU" |
| On-demand H200 141 GBper GPU hour | $8per minute | – | – | – | "ion below for details on dedicated training api pricing. on-demand pricing pay per gpu sec" |
| On-demand B200 180 GBper GPU hour | $13per minute | – | – | – | "per gpu second, with no extra charges for start-up times on demand deployments region-rest" |
| On-demand B300 288 GBper GPU hour | $15per minute | – | – | – | "per gpu second, with no extra charges for start-up times on demand deployments region-rest" |
| Enterprise | Quote | – | – | – | "Contact us for enterprise deployments with faster speeds, lower costs, and higher rate lim" |
Fine print
- Billing increment
- Per GPU second for on-demand deployments "Pay per GPU second, with no extra charges for start-up times"
- Billed beyond usage
- Usage above a Reserved Throughput reservation bills at standard Serverless rates "Usage above your reservation bills at standard Serverless rates, with no penalty."
- Contract
- Not stated
- Refunds
- Not stated
- Auto renew
- Not stated
- Price change notice
- Not stated
"Not stated" means we read their pricing page and terms and they don't say.
About the company
- Legal name
- Fireworks.ai, Inc.
- Online since
- Mar 11, 2020 Domain registration date
- Ships updates
- 7 in the last 12 months Last on Nov 7, 2026, from their changelog
- Founder
- Lin Qiao, Co-Founder, CEO
- Founder
- Benny Chen, Co-Founder
- Founder
- Dmytro Dzhulgakov, Co-Founder
- Founder
- Pawel Garbacki, Co-Founder
- Founder
- Dmytro Ivchenko, Co-Founder
- Founder
- James Reed, Co-Founder
- Founder
- Chenyu Zhao, Co-Founder
- Customers
- "Trusted by 10,000+ companies globally" their claim
Fireworks AI alternatives
Mistral AI$5.99 /moFree plan · checked Oct 29, 2026
Hugging Face$9 /moFree plan · checked Oct 29, 2026
Anthropic Claude$17 /moFree plan · checked Oct 29, 2026OPOpenAI$20 /moFree plan · checked Oct 29, 2026
Upstage$100 /moFree plan · checked Oct 29, 2026
Cohere$2,000 /moFree plan · checked Oct 29, 2026
AI21 Labs$0.20 /minutechecked Oct 29, 2026
Alibaba QwenQuoteFree plan · checked Oct 29, 2026Reviews elsewhere
We link to reviews instead of copying ratings: those sites don't allow reuse, and we don't test products ourselves.
Questions
Is Fireworks AI free?
There's no free plan. Free trial: Muse Glimmer 30B SFT free via UI and Serverless Training API until 11/30, no credit card required.
Does Fireworks AI train AI on your data?
No, according to Fireworks AI: "We do not use your prompts, training data, or API inputs to train or improve our AI models without your explicit opt-in."
Is Fireworks AI HIPAA compliant?
Fireworks AI says a HIPAA BAA is available.