Mistral Large 4 coding test: 2 hours of agentic work burned just €0.82 of quota

qtnx_ · x · 2026-11-19

Developer julesprv shared a hands-on test of Mistral Large 4: two hours of work on a personal eval harness with few or no errors, good and precise code, though not the fastest model.

Cost-wise, 2 hours of continuous high-reasoning usage consumed only €0.82 against a €255 usage cap, with no 5-hour rolling limit like competing plans. At €15/month, the author calls the quality/quantity/price ratio extremely compelling, and a screenshot shows the agent autonomously exploring a codebase and validating log formats.

Original post →

More from coding & agent

coding & agent channel →