Frontier Log

MistralModel

Mistral Large 3 675B

Made by
Mistral
Released
Announced in
Introducing Mistral 3mistral.ai
Event
Mistral 3
API record
/api/v1/models/mistral-large-3

Benchmark results

BenchmarkScoreSource check
GPQA Diamondbase model, before instruction tuning; 5-shot; no chain of thought43.9%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
instruct model; shots and chain of thought not stated · reported later, in Introducing Mistral Small 464.1%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
MMLU-Proinstruct model · reported later, in Introducing Mistral Small 480.9%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
LiveCodeBenchbase model, before instruction tuning; no chain of thought; problem window not stated34.4%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Sources

Reported by Mistral

Events

All models