Skip to main content

Fireworks AI API Pricing (2026)

Fireworks AI delivers production-grade inference for open-source models with industry-leading speed and competitive per-token pricing, including serverless and dedicated deployment options. Founded in 2022 and headquartered in San Francisco, CA, Fireworks AI offers 8 active models through their API. All pricing data below is sourced directly from the Fireworks AI pricing page and is updated regularly.

Prices verified Mar 10, 2026

Fireworks AI Model Pricing — All Models

All prices are in USD per 1 million tokens ($/M tokens). Input tokens are the text you send to the model; output tokens are the generated response.

ModelTierInputOutputCached InputBatch InputBatch Output
Llama 4 Maverick (Fireworks)
Meta's Llama 4 Maverick MoE model on Fireworks AI. A large open-source model with 131K context, delivering strong reasoning and coding performance at serverless pricing.
Mid$0.50/M$0.50/M$0.25/M$0.25/M$0.25/M
Llama 4 Scout (Fireworks)
Meta's Llama 4 Scout model on Fireworks AI. A lightweight MoE model with a 10M token context window and exceptional cost efficiency.
Budget$0.20/M$0.20/M$0.10/M$0.10/M$0.10/M
DeepSeek V3.1 (Fireworks)
DeepSeek V3.1 on Fireworks AI — one of the best open-weight models available, serving at production scale with Fireworks' optimized inference infrastructure.
Mid$0.56/M$1.68/M$0.28/M$0.28/M$0.84/M
Llama 3.3 70B Instruct (Fireworks)
Meta's Llama 3.3 70B Instruct on Fireworks AI. A high-quality 70B open-source model delivering near-frontier performance at a very competitive price.
Mid$0.90/M$0.90/M$0.45/M$0.45/M$0.45/M
Qwen3 VL 30B (Fireworks)
Alibaba's Qwen3 VL 30B vision-language model on Fireworks AI. Excels at image understanding, document OCR, and multimodal reasoning tasks.
Mid$0.15/M$0.60/M$0.08/M$0.08/M$0.30/M
GPT OSS 120B (Fireworks)
OpenAI's open-weight GPT OSS 120B model, served via Fireworks AI at production scale. Delivers strong reasoning and instruction-following at a mid-tier price.
Mid$0.15/M$0.60/M$0.08/M$0.08/M$0.30/M
GPT OSS 20B (Fireworks)
OpenAI's open-weight GPT OSS 20B model on Fireworks AI. A compact model offering very competitive inference pricing for cost-sensitive production workloads.
Budget$0.07/M$0.30/M$0.04/M$0.04/M$0.15/M
Kimi K2 Instruct (Fireworks)
Moonshot AI's Kimi K2 Instruct on Fireworks AI. A large MoE model optimized for agentic tasks, multi-step reasoning, and complex instruction following.
Mid$0.60/M$2.50/M$0.30/M$0.30/M$1.25/M

Prices last verified: 2026-03-10. Always confirm on the official pricing page before production use.

Fireworks AI Model Capabilities

Key technical capabilities across all Fireworks AI models. Use this table to identify which models support the features your application requires.

ModelContextMax OutputVisionFunctionsJSON ModeCachingBatch APIReasoning
Llama 4 Maverick (Fireworks)131,07216,384
Llama 4 Scout (Fireworks)10,000,00016,384
DeepSeek V3.1 (Fireworks)163,84016,384
Llama 3.3 70B Instruct (Fireworks)131,0728,192
Qwen3 VL 30B (Fireworks)262,1448,192
GPT OSS 120B (Fireworks)131,07216,384
GPT OSS 20B (Fireworks)131,07216,384
Kimi K2 Instruct (Fireworks)131,07216,384

When to Choose Fireworks AI

Cost-sensitive workloads

Use GPT OSS 20B (Fireworks) Fireworks AI's most affordable model at $0.07/M input / $0.30/M output. Estimated cost at 100K requests/month: $22.00.

Output-heavy workloads

Use Llama 4 Scout (Fireworks) for workloads where output tokens dominate (e.g., content generation). Output price: $0.20/M.

See detailed cost comparisons between Fireworks AI models and models from other providers, including side-by-side pricing tables and monthly cost projections.

Frequently Asked Questions: Fireworks AI API Pricing

What is the cheapest Fireworks AI model?

The cheapest Fireworks AI model by input token price is GPT OSS 20B (Fireworks), priced at $0.07/M for input tokens and $0.30/M for output tokens. It is best suited for high-volume, cost-sensitive workloads where speed and price matter most.

What is the most capable Fireworks AI model?

Llama 4 Maverick (Fireworks) is Fireworks AI's most capable model, classified as a mid-tier model. It supports a context window of 131,072 tokens and is priced at $0.50/M input / $0.50/M output per million tokens.

How does Fireworks AI API pricing work?

Fireworks AI charges per token consumed — separately for input tokens (the text you send) and output tokens (the text the model generates). Prices are listed in USD per 1 million tokens ($/M). Fireworks AI was founded in 2022 and is headquartered in San Francisco, CA. All pricing shown here is sourced from the official Fireworks AI pricing page.

Does Fireworks AI offer batch API pricing?

Yes — 8 Fireworks AI models support batch processing at reduced rates: Llama 4 Maverick (Fireworks), Llama 4 Scout (Fireworks), DeepSeek V3.1 (Fireworks), and others. Batch API is ideal for asynchronous workloads like bulk data extraction, document processing, or offline classification pipelines. Batch pricing is shown in the pricing table above.

Which Fireworks AI models support prompt caching?

8 Fireworks AI models support prompt caching: Llama 4 Maverick (Fireworks), Llama 4 Scout (Fireworks), DeepSeek V3.1 (Fireworks), Llama 3.3 70B Instruct (Fireworks), Qwen3 VL 30B (Fireworks), GPT OSS 120B (Fireworks), GPT OSS 20B (Fireworks), Kimi K2 Instruct (Fireworks). Prompt caching stores repeated context (system prompts, knowledge base content) in memory and applies a discount to cached input tokens, making it highly effective for applications that reuse the same context across many requests.

Which Fireworks AI models support vision / image inputs?

3 Fireworks AI models support vision (image input): Llama 4 Maverick (Fireworks), Llama 4 Scout (Fireworks), Qwen3 VL 30B (Fireworks). These models can analyze images, charts, screenshots, and documents alongside text prompts.

Fireworks AI pricing data last verified: 2026-03-10. Prices may change — verify on the official Fireworks AI pricing page before production use.