Fireworks AI - AI Tools Nav

What it is for

Fireworks AI is a serverless (and dedicated-GPU) inference catalog: browse the Model Library, then call a model billed per 1M tokens. Use it when the job is one HTTP call to an open model with a printed row.

What makes it unique: the cover is the live fireworks.ai/models library — search, Serverless / Embeddings / LLM filters, and cards such as Ember-1 and DeepSeek V4.1 Flash.

Key features

Model Library — search plus Serverless / Embeddings / Rerank / LLM / Router / Vision chips (cover) on fireworks.ai/models.
Headline cards — Ember-1, DeepSeek V4.1 Flash, FireRouter, and other NEW rows on the same grid.
Serverless inference — one HTTP call billed on the input / cached / output triple (Serverless Pricing).
On-demand GPUs — dedicated H100 / H200 / B200 boxes on fireworks.ai/pricing.
Fine-tune / LoRA SFT — training SKUs sit next to the serverless catalog.
Embeddings and rerank — separate chips and tables on the same library.

Who it's for

Best fit: developers who will pick a library card and copy the Standard $/1M triple.

Pricing

Verified 2026-10-22 (Asia/Shanghai) from Serverless Pricing and fireworks.ai/pricing. Prices are USD per 1M tokens (input / cached / output). Batch is 50% of serverless on input and output.

Standard examples:

DeepSeek V4.1 Flash — $0.30 / $0.006 / $1.20
GLM 5.3 Flash — $0.15 / $0.03 / $0.50
GPT-OSS 120B — $0.15 / $0.015 / $0.60
Ember-1 — $3.00 / $0.30 / $15.00
NVIDIA Nemotron 3.5 Lightning 30B A3B — $0.05 / $0.01 / $0.20

On-demand GPUs (per hour): H100 / H200 $8.00; B200 $13.00; B300 $15.00; GB300 $20.00.

Embeddings: up to 150M $0.008 / 1M input; 150M–350M $0.016; Qwen3 8B $0.10.

Limitations / costs

You are billed per token (or per GPU hour).
Priority and Fast are other columns on the same docs table; some rows leave Priority blank.
US-only rows carry a listed premium (DeepSeek V4.1 Flash US $0.45 / $0.009 / $1.80).
Size-based fallbacks apply to models outside the headline table ($0.10–$1.20 / 1M).

Alternatives

Together AI — hosted open-model catalog with printed $/1M cards.
Groq — GroqCloud Supported Models table with GPT-OSS $0.15 / $0.60.
OpenRouter — one key across many providers.

Official tutorial

Written path: Model Library plus Serverless Pricing.

Website

https://fireworks.ai

Source

Catalog row curated for AI Tools Nav. Cover is the live fireworks.ai/models library (search + Ember-1 / DeepSeek V4.1 Flash cards). Dollars checked on docs.fireworks.ai/serverless/pricing and fireworks.ai/pricing (Flash $0.30 / $0.006 / $1.20; H100 $8/hr).

Last verified

2026-10-22 (Asia/Shanghai) editorial check. Only printed Standard triples and GPU-hour rows were copied.