Mistral Large 3 / Ministral 3 14B / Ministral 3 8B / Ministral 3 3B

heading: Mistral Large 3: A state-of-the-art open model · quote_snippet: Mistral Large 3 debuts at #2 in the OSS non-reasoning models category (#6 amongst OSS models overall)

{
  "harness": null,
  "tools": null,
  "shots": null,
  "reasoning_effort": "non-reasoning category",
  "temperature": null,
  "top_p": null,
  "token_budget": null,
  "turn_limit": null,
  "time_limit": null,
  "run_count": null,
  "aggregation": "leaderboard",
  "judge": null
}

LMArena leaderboard rank cited by Mistral; Elo value not printed in text and the accompanying LMArena chart is an image (not machine-read). Third-party leaderboard quoted by vendor, so not vendor self-report and excluded from the official-citation count per goal.md 12.3.

打开官方来源

heading: Ministral 3: State-of-the-art intelligence at the edge · quote_snippet: for instance 85% on AIME '25 with our 14B variant

{
  "harness": null,
  "tools": null,
  "shots": null,
  "reasoning_effort": "reasoning variant ('can think longer')",
  "temperature": null,
  "top_p": null,
  "token_budget": null,
  "turn_limit": null,
  "time_limit": null,
  "run_count": null,
  "aggregation": null,
  "judge": null
}

new-benchmark: aime-25 not yet in data/benchmarks/. AIME 2025 edition is a different task set from the existing aime24 (AIME 2024) entry, so a distinct id is used rather than a variant of aime24.

打开官方来源