Mistral Large 4 Enters Agent Arena Top 50, Demonstrating Significant Agentic Capability Gains

Mistral AI's new Large 4 model has claimed the 43rd position on Agent Arena's leaderboard, marking a substantial improvement over its predecessor in real-world agentic task performance. The 1 trillion-parameter open-weight model, which entered the benchmark in early October 2026, recorded a 6.56% net improvement score across more than 5,000 real-world agentic sessions — placing it 11 rankings above Mistral Medium 3.5, the company's prior generation agent-oriented variant.