Key Info

Mistral Large 4 is positioned as a general-purpose agent model that gathers information and produces finished deliverables across complex workflows. On AutomationBench, it was evaluated on 657 business workflows spanning simulated workplace apps such as Gmail, Google Sheets, Slack and Salesforce, finishing ahead of Kimi K3, MiMo-V2.6-Pro and DeepSeek V4 Pro.

Highlights

  • Mistral Large 4 targets end-to-end agent workflows, not just single-turn responses
  • AutomationBench covers 657 business workflows across simulated workplace applications
  • Competing models it finished ahead of: Kimi K3, MiMo-V2.6-Pro, DeepSeek V4 Pro
  • On Harvey's Legal Agent Benchmark (LAB), Mistral Large 4 ranks as the #1 open-source model
  • Common work apps tested include Gmail, Google Sheets, Slack and Salesforce