Skip to content
aiqbench

Anthropic · model passport

Claude 3 7 Sonnet

claude-3-7-sonnet
  • Cloud only
  • Proprietary licence
  • Released 2025-02-24(curated)
17/27axes covered
Overall · LMArena
1,388#165 of 396
Best percentile
86.1MMLU-Pro (Artificial Analysis)
Cohort standing
52.3median 44.4

What the data says

Benchmarked on 17 of 27 tracked axes. Every figure below is the model's own published result on that board.

Strongest axes

  • 86.1 pct — MMLU-Pro (Artificial Analysis) (83.7%)

Uneven: 70.5 percentile points between its best and worst axis — 86.1 pct on MMLU-Pro (Artificial Analysis) versus 15.6 pct on SWE-bench Verified.

  • MMLU-Pro (Artificial Analysis)86.1 pct83.7%Verify →
  • LiveBench84.2 pct70.6%Verify →
  • SWE-bench Verified (live board)78.5 pct70.2%Verify →
  • Instruction following70.7 pct1,411Verify →
  • Coding64.2 pct1,452Verify →
  • Hard prompts62.1 pct1,418Verify →
  • GPQA Diamond59.7 pct78.5%Verify →
  • Overall58.6 pct1,388Verify →
  • LiveCodeBench (Artificial Analysis)55 pct47.3%Verify →
  • IFBench (Artificial Analysis)54.9 pct48.3%Verify →
  • Vision Arena44 pct1,193Verify →
  • SimpleBench43.7 pct46.4%Verify →
  • AIME (math)42.5 pct57.8%Verify →
  • FrontierMath29 pct4.1%Verify →
  • Humanity's Last Exam28.3 pct8.0%Verify →
  • ARC-AGI-218 pct0.9%Verify →
  • SWE-bench Verified15.6 pct61.0%Verify →
  • WebDev ArenaNot benchmarked
  • Factuality ArenaNot benchmarked
  • Search ArenaNot benchmarked
  • Document ArenaNot benchmarked
  • Agent ArenaNot benchmarked
  • SimpleQA VerifiedNot benchmarked
  • BFCL v4 (function calling)Not benchmarked
  • Terminal-Bench 4.0Not benchmarked
  • HHEM Hallucination (Vectara)Not benchmarked
  • Terminal-BenchNot benchmarked

Percentiles are the model's standing within each board's own population. Axes marked "not benchmarked" have no published result for this model — that is an absence of data, not a low score.

  • claude-3-7-sonnet-20250219-thinking-32k stands at 52.3 (raw mean 52.7) — measured on 17 of 27 axes.

    standingtemplateVerify →

  • claude-3-7-sonnet-20250219-thinking-32k stands at 52.3 versus a cohort median of 44.4 across 147 models released within 180 days — above the peer median.

    cohorttemplateVerify →

Every claim is a computed fact from the current board data, phrased by a low-temperature model call and rejected by a validation gate if any number is not in the source row. How the leaderboards work →