Mistral AI released Mistral Medium 3.5, its first flagship merged model: a dense 128B parameter system that combines instruction-following, reasoning, and coding in a single set of weights with a 256k context window. The model scores 77.6% on SWE-Bench Verified and 91.4 on the τ³-Telecom agentic benchmark, ahead of Devstral 2 and Qwen3.5 397B A17B. Weights are released on Hugging Face under a modified MIT license and can run on as few as four GPUs for self-hosting; API pricing is $1.50 input / $7.50 output per million tokens. The model replaces Devstral 2 as the default in Mistral's Vibe coding agent and powers a new Work mode in Le Chat for long-horizon tasks.