Mistral announces Mistral Large 4, scoring 38 on Artificial Analysis Intelligence Index; weights not yet released
2026-10-16
Mistral announced Mistral Large 4. Per Nathan Lambert's Interconnects analysis, Artificial Analysis scored it 38 on its Intelligence Index — well below the current closed frontier (~58) and behind current open-weights leaders GLM-5.3 and Kimi K3. As of the report, Large 4's weights had not yet been released.
Significance 3: A new flagship announcement from a tracked lab with an independent (Artificial Analysis) benchmark score, but it is not yet a shipped open-weights release and the score trails both the closed frontier and current open leaders, keeping this below a verdict-moving threshold.
Implications · machine-drafted, not owner judgment
A roughly 20-point Intelligence Index gap to the closed frontier for the leading non-Chinese open-weights lab's new flagship is a mixed signal for Mistral's open-weights-through-2027 commitment that underpinned its recent multi-billion-euro raises, even as Meta's Muse Spark Open and Reflection's Beam narrow the US open-weights gap elsewhere this same week. If Mistral ships Large 4 open and the score holds, it weakens the one-generation-behind thesis specifically for the European/sovereign-AI segment.
- Whether and when Mistral actually releases Large 4's weights
- Independent confirmation or revision of the 38 Intelligence Index score once broader access is available
Sources
- primary Introducing Mistral Large 4 retrieved 2026-10-17
- curator Meta shakes up the frontiers, open and closed retrieved 2026-10-17
Affected axes
Hard reasoning