Mistral Small 4 launched in mid-March 2026 as an efficient 119B MoE hybrid model unifying instruction following, reasoning, multimodal input, and agentic coding under an Apache 2.0 license, with a 256k context window and only ~6B active parameters per token. Since then, the company has shifted focus to larger releases, including Mistral Medium 3.5 in April and the 1T-parameter Mistral Large 4 entering public preview on October 6, 2026, with open weights expected later that month. No announcements or credible reports have surfaced for a Small 4.1+ update through early October, consistent with Mistral’s pattern of spacing major Small-family refreshes by several months amid competition from open-weight models like Qwen and GLM series. Traders are watching for any post-Large 4 signals on smaller-model roadmaps or developer conferences that could accelerate the next iteration.