Mistral Coding Model Benchmarks Questioned as License Terms Draw Scrutiny

Mistral reported its coding models scoring 19.7% on FrontierCode 2.0 and 41% on Terminal-Bench 4.0, but Julien Simon noted the 41% figure only holds within Mistral's own coding agent. The models are commercially usable, though selling them as a paid API requires a separate agreement.

2026-11-01 ~ 2026-11-01 · 4 related posts

• Mistral Large 4's Terminal-Bench 41% only holds inside Mistral's own agent — julsimon · 2026-11-01
• Mistral License in Short: Free to Use and Ship, Paid API Hosting Needs a Deal — AIFlow_ML · 2026-11-01
• Mistral Reports 19.7% on FrontierCode 2.0, 41% on Terminal-Bench — Their Own Harness — AIFlow_ML · 2026-11-01
• Mistral coding model scores 19.7% FrontierCode, 41% Terminal-Bench — but not for 24GB cards — AIFlow_ML · 2026-11-01