Unverified50% confidenceBenchmarkExact time
In the Harvey legal agent evaluation, Mistral Large 4 surpassed all open-source models.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/30/2026
First Seen
Valid until: 11/13/2026
Sources
Mistral Large 4 Released: 1T-Parameter MoE Tops Western Open-Source Leaderboard
bilibili一只AI风向标10/6/2026
Related Entities
Related Claims
Partially VerifiedMistral Large 4 adopts a sparse Mixture of Experts (MoE) architecture, with 49 billion (49B) parameters actually activated during each inference.68% similarUnverifiedMistral Large 4 performs roughly on par with GRM 5.3 on financial Agents and is the strongest model in legal Agents and data-grounded capabilities.68% similarVerifiedMistral released Mistral Large 4, codenamed LeChunk, a trillion-parameter, natively multimodal MoE model with approximately 49 billion active parameters.67% similarUnverifiedMistral Large 4 scored approximately 62% on the real-world software engineering benchmark DeepSwee.67% similarUnverifiedIn deep software engineering tests, Mistral Large 4 is inferior to Kimi K3, with a clear gap.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1115000API
curl https://kongchang.com/api/v1/knowledge/claims/1115000MCP
get_claim(id=1115000)