Mistral Medium 3.5, the current 128B dense flagship unifying instruction following, reasoning, and coding under a modified MIT open-weight license with 256K context, launched in late April 2026 and remains the production default powering Vibe. No 3.6+ iteration has shipped since, with Mistral instead prioritizing specialist drops and the October 6 preview of its much larger 1T-parameter Large 4 MoE model (weights slated for end-of-month release). Trader sentiment reflects the lab’s irregular cadence after consolidating prior Medium, Magistral, and Devstral lines, alongside competitive pressure from faster-moving closed models at Anthropic and OpenAI plus strong open-weight rivals from Chinese labs. Key near-term signals include any post-Large 4 roadmap updates or benchmark gains on agentic and SWE-Bench tasks that could accelerate a Medium follow-up.