by Mistral AI

Mistral Medium 3.5.

multimodal closed 262K ctx
Cheapest input
$1.5/M
on Mistral La Plateforme
Cheapest output
$7.5/M
on Mistral La Plateforme
Hosted equiv.
~$2.7/hr
@ 100 tok/s on Mistral La Plateforme
Description

About Mistral Medium 3.5.

Mistral Medium 3.5 is a dense 128-billion-parameter model that handles instructions, reasoning and coding in one set of weights. It takes text and images, returns text, and has a 256k context, with reasoning effort set per request. Mistral uses it in place of Medium 3.1 and Magistral in Le Chat, and Devstral 2 in its Vibe coding agent. The Modified MIT License has exceptions for companies with large revenue. An early Transformers config hurt long-context results, and GGUF files built from it before the fix are affected too.

Context window

How much it can remember.

262K tokens ≈ 196,608 English words
4K 32K 128K 1M
Capabilities

What it can do.

· Vision input
· Audio input
· Video input
· Function calling
· Tool use
· JSON mode
✓ Streaming
· Fine-tuning