I tested Mistral Small 4 in an Agentic Workflow
#7
by zacksiri - opened
So this model failed one of the tests but passed the others. It's very odd, maybe it's something to do with MoE model.
https://upmaru.com/llm-tests/simple-tama-agentic-workflow-q1-2026/mistral-small-4
I also tried enabling reasoning_effort "high" it did not improve the situation.
zacksiri changed discussion status to closed