128B (dense)
256K tokens
Modified MIT License
Official open-weight flagship merged Mistral release that unifies instruction-following, reasoning, coding, and vision in a single dense 128B model. Mistral positions Medium 3.5 as the successor to both Mistral Medium 3.1 and Magistral in Le Chat, and as the replacement for Devstral 2 in Vibe. Release materials emphasize configurable reasoning effort per request, multimodal text-plus-image input with text output, strong system-prompt adherence, native function calling and JSON output, and self-hosting on as few as four GPUs. The official model card reports 77.6% on SWE-Bench Verified and 91.4 on tau-cubed Telecom, with the model aimed at long-horizon coding, agentic tool use, and structured productivity workflows. Mistral also released a companion EAGLE head for faster speculative decoding in vLLM deployments.