Mistral Medium 3.5
Mistral AI · Multimodal model
Mistral Medium 3.5 is a 128B dense model that combines chat, reasoning and coding in one set of open weights. It is the default model in Le Chat and replaced the Devstral coding models in Mistral's Vibe coding agent.
Current Open weights
ReleasedApr 28, 2026
Context window256K tokens
Input → outputtext, image → text
API price$1.5 per million input tokens and $7.5 per million output tokens
Consumer accessDefault model in Le Chat; remote coding agents and Work mode (preview) on Pro ($14.99/month), Team and Enterprise plans
Best for
Coding agents and everyday assistant work where you want an open-weight model that can run on a few GPUs.
Overview
Mistral Medium 3.5 is a 128B dense model that combines chat, reasoning and coding in one set of open weights. It is the default model in Le Chat and replaced the Devstral coding models in Mistral's Vibe coding agent.
Strengths and limits
- 77.6% on SWE-Bench Verified per Mistral
- Dense 128B model, self-hostable on as few as four GPUs
- Per-request reasoning effort; powers Le Chat and Mistral Vibe
- Modified MIT license, not plain Apache 2.0
- Output price ($7.5 / 1M) is higher than Mistral Large 4's list price
Published benchmark scores
SWE-Bench Verified77.6%
Source: mistral.ai
τ³-Telecom91.4
Source: mistral.ai
Scores are as published by the source; settings (reasoning effort, tools, attempts) differ between labs, so compare with care. How to read benchmarks →