# Fireworks AI — AI Compare Ninja

> US inference platform built by the creators of PyTorch for fast, cost-efficient deployment of open and proprietary AI models.

## Overview
- **Category:** Development & Code (Agent & RAG frameworks)
- **Status:** Active · checked - Active.
- **Paid from:** H100/H200 8 $, B200 13 $, B300 15 $, GB300 20 $ per GPU hour / Custom (Fireworks AI)
- **Free access:** Free plan
- **Website:** https://fireworks.ai/
- **Canonical URL:** https://www.aicompare.ninja/en/tool/fireworks-ai

## What it does
- US inference platform by PyTorch founders for fast, cost-efficient AI execution
- Industry-leading inference throughput and low latency
- Wide library of open-source models and Fireworks proprietary models
- Serverless API with no upfront cost for immediate use
- Compound AI system support for complex agent pipelines
- Fine-tuning and dedicated deployments available

## In brief
Fireworks AI is a US inference platform founded by PyTorch founders for fast, cost-efficient AI model execution. Serverless API and dedicated deployments enable scalable integration without upfront costs.

The platform offers open-source models, Fireworks proprietary models, and compound AI system support for complex agent pipelines. Fine-tuning is available for customised model variants.


## Limits
On-demand deployments are billed per GPU second with no startup-time charge; region-restricted deployments carry a 1.5x premium.

## Pricing tiers
| Tier | Price | Period | What you get |
| --- | --- | --- | --- |
| Fireworks AI | H100/H200 8 $, B200 13 $, B300 15 $, GB300 20 $ per GPU hour | Custom | H100/H200 $8, B200 $13, B300 $15, GB300 $20 per GPU hour |

## Sources
All details regarding plans, limits, and features are based on official vendor documentation and the primary sources verified below.

- https://fireworks.ai/
- https://fireworks.ai/pricing

---
*Source: AI Compare Ninja (2026) - https://www.aicompare.ninja/en/tool/fireworks-ai*