# Mistral Medium 3

> Frontier-class performance at mid-tier cost — Mistral AI's enterprise workhorse.

- Maker: Mistral AI
- Line: Mistral Medium
- Canonical page: https://ai-tldr.dev/models/mistral-medium-3/
- Updated: 2026-06-18

## In short

Mistral Medium 3 is an API-only model from Mistral AI. Mistral AI released it on 7 May 2025. It has a 128K-token context window and costs $0.40 per million input tokens and $2.00 per million output tokens. You can use it through Mistral AI (La Plateforme), OpenRouter and Amazon SageMaker.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 128K | [1] |
| Input price | $0.40 / 1M tokens | [2] |
| Output price | $2.00 / 1M tokens | [2] |
| Licence | Proprietary · API only | [1] |
| Released | 7 May 2025 | [1] |
| Knowledge cutoff | Not disclosed | [1] |

## Sources

1. [Mistral Medium 3 model card (docs)](https://docs.mistral.ai/models/model-cards/mistral-medium-3-25-05)
2. [Mistral AI — Medium 3 announcement](https://mistral.ai/news/mistral-medium-3/)

## Overview

Mistral Medium 3 (model ID mistral-medium-2505) is a mid-tier large language model released by Mistral AI on May 7, 2025 under the tagline "Medium is the new large." Unlike many of Mistral's earlier models, Mistral Medium 3 is proprietary rather than open-weight — it is served through Mistral's API and cloud partners, and was positioned as a cost-efficient enterprise model that delivers frontier-level quality at roughly an order of magnitude lower cost than traditional large models.

Mistral AI claimed Mistral Medium 3 performs at or above 90% of Anthropic's Claude 3.7 Sonnet across benchmarks, while standing out in coding and STEM tasks and surpassing leading open models such as Llama 4 Maverick and enterprise models such as Cohere Command A. The model is multimodal: it accepts text and image input, handles document understanding (with OCR and strong DocVQA results), and supports a 128K-token (131K) context window.

Mistral Medium 3 was designed for flexible deployment — it can run on any cloud or in self-hosted environments of four GPUs and above. It launched on Mistral's La Plateforme and Amazon SageMaker, with availability on IBM watsonx, NVIDIA NIM, Azure AI Foundry, and Google Cloud Vertex following. It was later succeeded by Mistral Medium 3.1 (August 2025) and Mistral Medium 3.5 (April 2026), and is being deprecated in 2026.

## Benchmarks

- HumanEval (0-shot): 92.1% ([source](https://apidog.com/blog/mistral-medium-3/))
- Math500 Instruct (0-shot): 91% ([source](https://apidog.com/blog/mistral-medium-3/))
- MMLU Pro (5-shot CoT): 77.2% ([source](https://apidog.com/blog/mistral-medium-3/))
- GPQA Diamond (5-shot CoT): 57.1% ([source](https://apidog.com/blog/mistral-medium-3/))
- ArenaHard (0-shot): 97.1% ([source](https://apidog.com/blog/mistral-medium-3/))
- IFEval (0-shot): 89.4% ([source](https://apidog.com/blog/mistral-medium-3/))
- MMMU (0-shot): 66.1% ([source](https://apidog.com/blog/mistral-medium-3/))
- DocVQA (0-shot): 95.3% ([source](https://apidog.com/blog/mistral-medium-3/))
- RULER 128K: 90.2% ([source](https://apidog.com/blog/mistral-medium-3/))
- LiveCodeBench v6 (0-shot): 30.3% ([source](https://apidog.com/blog/mistral-medium-3/))

## Pricing

- Input: $0.40 / 1M tokens per 1M tokens
- Output: $2.00 / 1M tokens per 1M tokens

Source: https://mistral.ai/news/mistral-medium-3/

## Strengths

- Strong coding performance — HumanEval 92.1%, close to much larger and slower competitors
- Competitive STEM and reasoning — Math500 91.0% and MMLU Pro 77.2%
- Multimodal: handles images and documents (DocVQA 95.3%, MMMU 66.1%)
- Very low cost for its tier at $0.40 input / $2.00 output per 1M tokens
- Flexible deployment — any cloud, self-hosted, or on-prem from four GPUs up
- Reliable instruction following (IFEval 89.4%) and long-context retrieval (RULER 128K 90.2%)

## Best for

- Cost-sensitive enterprise coding assistants and code generation
- STEM, math, and technical reasoning workloads
- Document and image understanding (OCR, DocVQA-style extraction)
- Customer-facing assistants that need quality at controlled cost
- On-premise or self-hosted deployments where data must stay in-house
- Long-context tasks up to 128K tokens

## How to access

- [Mistral AI (La Plateforme)](https://docs.mistral.ai/models/model-cards/mistral-medium-3-25-05) — model id `mistral-medium-2505`
- [OpenRouter](https://openrouter.ai/mistralai/mistral-medium-3) — model id `mistralai/mistral-medium-3`
- [Amazon SageMaker](https://mistral.ai/news/mistral-medium-3/)

## Mistral Medium versions

- [Mistral Medium 3.5](https://ai-tldr.dev/models/mistral-medium-3-5/) (28 April 2026)
- [Mistral Medium 3.1](https://ai-tldr.dev/models/mistral-medium-3-1/) (12 August 2025)
- Mistral Medium 3 (7 May 2025) — this page
- [Mistral Medium (2023)](https://ai-tldr.dev/models/mistral-medium-legacy/) (December 2023)

## FAQ

### Is Mistral Medium 3 open weights?

No. Unlike many earlier Mistral models, Mistral Medium 3 is proprietary. It is offered through Mistral's API and cloud partners rather than as a downloadable open-weight model. (Mistral later released Mistral Medium 3.5 as open weights, but that is a different, newer model.)

### How much does Mistral Medium 3 cost?

Mistral Medium 3 is priced at $0.40 per million input tokens and $2.00 per million output tokens — positioned as roughly an order of magnitude cheaper than traditional large models, which is central to its enterprise pitch.

### What is Mistral Medium 3's context window and can it process images?

Mistral Medium 3 has a 128K-token (131K) context window and is multimodal: it accepts text and image input and handles document understanding, scoring 95.3% on DocVQA and 66.1% on MMMU.

### How does Mistral Medium 3 compare to Claude 3.7 Sonnet?

Mistral AI claimed Mistral Medium 3 performs at or above 90% of Claude 3.7 Sonnet across benchmarks while costing far less, and it is especially competitive in coding (HumanEval 92.1%) and STEM tasks (Math500 91.0%, MMLU Pro 77.2%).

## Links

- [Mistral AI — Medium 3 announcement](https://mistral.ai/news/mistral-medium-3/)
- [Mistral Medium 3 model card (docs)](https://docs.mistral.ai/models/model-cards/mistral-medium-3-25-05)
- [Mistral API documentation](https://docs.mistral.ai/models/overview)
- [OpenRouter — Mistral Medium 3](https://openrouter.ai/mistralai/mistral-medium-3)
- [Artificial Analysis — Mistral Medium 3](https://artificialanalysis.ai/models/mistral-medium-3)
- [Benchmark scores reference (apidog)](https://apidog.com/blog/mistral-medium-3/)
- [SiliconANGLE — launch coverage](https://siliconangle.com/2025/05/07/mistral-ai-adds-medium-3-family-models-claiming-low-cost-high-performance/)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/mistral-medium-3/