Mistral Medium 3.5 is a dense 128B instruction-following model from Mistral AI.
- Context window: 262,144 tokens
- Input: text, image, file
- Output: text
- Pricing: $1.5/M input tokens, $7.5/M output tokens
Community sentiment
Negative (-20) — 1 positive, 2 negative, 3 neutral across 6 community samples from X and the web, as of 2026-10-02.
Users note Mistral Medium 3.5's frontier-class capabilities (dense 128B architecture, very long 256K context, configurable reasoning and image understanding) and its open-weight / downloadable availability. Criticisms focus on low benchmark scores relative to competitors and complaints about it being particularly expensive compared with similar open-weight models.
- “Is Mistral still releasing anything? On Artificial Analysis Mistral Medium 3.5 scores 14. DeepSeek V4.1 Flash is 39 and GLM 5.3 is 45. Europe's main open lab is a whole tier behind the Chinese ones right now. https://t.co/2l77fiHNSh” — X
- “Mistral Medium 3.5 vs Muse Spark 1.3 - DocsBot — 6 days ago · Mistral Medium 3.5 is Mistral's frontier-class open. It is a dense 128B model with a 256K context window, configurable reasoning effort, image understanding, ...” — the web
- “Mistral AI raises €3B led by Samsung: how sovereign is it? — ## Mistral Medium 3.5: price and benchmarks Their summary also says it is "particularly expensive when comparing to other open weight models of similar size". Dev Twitter was less polite. Theo [wrote](https://x.com/theo” — the web
- “The open-weight ones you can download and run offline (via Hugging Face, Ollama, LM Studio etc.): DeepSeek V4.1 Flash, all Qwen3.8 variants, GLM-5.3 series, Kimi K3, Hy-4 Preview/Hy-3, MiniMax-M3, Nemotron 3 Ultra & 3.5 Lightning, Mistral Medium 3.5, Muse Glimmer, Inkling, and K2” — X
View on OpenRouter.
Model data sourced from OpenRouter.