Fireworks AI
CuratedActiveFast inference and fine-tuning platform with per-second on-demand GPU deployments
From
—
GPU from
$8.00/GPU-hr
Locations
0
countries
Data centers
0
sites
Networks (ASN)
0
ASNs
Plans & prices
5
plans
Overview
Fireworks AI, founded in 2022 by former PyTorch engineers at Meta, serves open models through a serverless token API and lets customers run on-demand deployments on dedicated GPUs (H100 to GB300) billed per GPU-second, plus fine-tuning and training APIs.
Plans & prices
GPU instances5
| Instance | Config | per GPU-hr | Total /hr | Spot | Locations | Source |
|---|---|---|---|---|---|---|
On-demand deployment, H200 141GB — per GPU, billed per second | 1× H200 | $8.00 | $8.00 | — | Verified 3 days ago | |
On-demand deployment, H100 80GB — per GPU, billed per second | 1× H100 SXM | $8.00 | $8.00 | — | Verified 3 days ago | |
On-demand deployment, B200 180GB — per GPU, billed per second | 1× B200 | $13.00 | $13.00 | — | Verified 3 days ago | |
On-demand deployment, B300 288GB — per GPU, billed per second | 1× B300 (Blackwell Ultra) | $15.00 | $15.00 | — | Verified 3 days ago | |
On-demand deployment, GB300 288GB — per GPU, billed per second | 1× GB300 (Blackwell Ultra) | $20.00 | $20.00 | — | Verified 3 days ago |
GPU models
Similar providers
Together AIOpen-model inference API plus self-serve and reserved NVIDIA GPU clusters
falGenerative media platform with pay-per-use model APIs and serverless GPUs
CirrascaleSan Diego AI cloud with NVIDIA, AMD, Qualcomm and Tenstorrent systems on term pricing
Hugging FaceAI model hub selling GPU Spaces, dedicated Inference Endpoints and routed inference
DeepInfraLow-cost model inference API plus dedicated A100-to-B300 GPUs billed by the minute
BasetenInference platform for dedicated model deployments and model APIs, billed per minute
GMI CloudNVIDIA Cloud Partner renting H100/H200/B200 and GB200 from its own data centers
GroqInference neocloud built on its own LPU accelerators, serving models via GroqCloud