Mistral AI

MMistral Medium 3OPEN

Mistral Medium 3 (Mistral AI): live API pricing, context window and the best open-source alternatives, tracked daily by OpenSourceAI.tech.

Context window
131K
tokens
Input price
$0.4
per M tokens
Output price
$2
per M tokens
Provider
Mistral AI

Prices update automatically — checked hourly against provider list prices.

See model comparisons → Compare all model prices

Benchmarks & performance

Independent benchmark scores for Mistral Medium 3, measured by Artificial Analysis. Higher is better.

2.6×
More intelligence per dollar than Claude Opus 5.
For the same budget, Mistral Medium 3 delivers 2.6 times more capability. Open weights also mean you can self-host it and pay nothing per token.
Intelligence index12.5
Math index30.3
GPQA57.8%
MMLU-Pro76%
Humanity's Last Exam4.3%
Long Context Reasoning28%
LiveCodeBench40%
SciCode33.1%
MATH-50090.7%
AIME44%
AIME 202530.3%
IFBench39.3%
τ²-Bench24.3%
Terminal-Bench Hard3.8%
💰 Blended price$0.8 / 1M tokens
📈 Value15.6 intelligence points per $
Benchmark data by Artificial Analysis

About this model

Mistral Medium 3 is an open-weight AI model by Mistral AI. You can download and self-host it for free; the prices below are hosted-API list prices, tracked hourly, for when you prefer convenience over self-hosting.

Frequently asked questions

What is Mistral Medium 3?

Mistral Medium 3 is an AI language model from Mistral AI. It is open-weight: you can download it and run it on your own hardware, for free. It scores 12.5 on the Artificial Analysis intelligence index.

Is Mistral Medium 3 free?

The weights are free and open — you can self-host Mistral Medium 3 and pay nothing per token. If you prefer a hosted API, list prices are $0.4 per million input tokens and $2 per million output tokens.

What is Mistral Medium 3 good at?

Independent benchmarks from Artificial Analysis give it GPQA 57.8%, MMLU-Pro 76%, Humanity's Last Exam 4.3%, Long Context Reasoning 28%, LiveCodeBench 40%, SciCode 33.1%, MATH-500 90.7%, AIME 44%, AIME 2025 30.3%, IFBench 39.3%, τ²-Bench 24.3%, Terminal-Bench Hard 3.8%. It is particularly used for mathematical reasoning.

Can I self-host Mistral Medium 3?

Yes. Mistral Medium 3 has open weights, so you can download it and run it on your own GPU or server with tools like Ollama, vLLM or llama.cpp — with no per-token cost.

Open-source alternatives

Depending on your workload, an open-weight model that you can self-host may cut costs dramatically or remove per-token pricing entirely. Our comparison pages put this model side by side with the strongest open-source contenders.

Mistral NemoMistral AIMistral Small 3.2 24BMistral AIMistral Small 4Mistral AIMistral Medium 3.5Mistral AIMistral Small 3Mistral AIMistral Large 3 2512Mistral AI