Share promptsSign inCreate team library
The Chat Judge verdicts

Which AI is best at what?

Public benchmark results for 19 tests across 9 fields, charted side by side. Every score links to the leaderboard it came from.

Showing Moonshot's best-ranked model on each benchmark.

1 benchmarkUpdated 19 Oct 2026Overall workOne number for agentic work across finance, coding, law and tax.
  • Vals Index Kimi K3 #18 · 52.95%
See charts 3 benchmarksUpdated 19 Oct 2026LegalDrafting, extraction, legal reasoning and US legal research.
  • Legal Benchmarks leaderboard Kimi K3 #7 · 80.0%
  • LegalBench Kimi K3 #16 · 86.21%
  • Legal Research Bench Kimi K3 #16 · 46.15%
See charts 3 benchmarksUpdated 21 Oct 2026Health & medicalPhysician-graded health conversations, medical coding and exam questions.
  • HealthBench (independent run) Kimi K3 (max) #9 · 54.7%
  • MedCode Kimi K3 #23 · 49.36%
  • MedQA Kimi K2.5 #17 · 94.37%
See charts 2 benchmarksUpdated 19 Oct 2026FinanceAnalyst research on SEC filings and long credit agreements.
  • Finance Agent Kimi K3 #20 · 60.01%
  • CorpFin v2 Kimi K3 #3 · 71.56%
See charts 2 benchmarksUpdated 1 Sep 2026CodingFixing real GitHub issues and solving fresh programming problems.
  • SWE-bench Verified Kimi K3 #8 · 93.4%
  • LiveCodeBench Kimi K3 #18 · 87.19%
See charts 2 benchmarksUpdated 19 Oct 2026Agents & tool useMulti-step tasks an agent completes alone in a terminal.
  • Terminal-Bench 2.1 Kimi K3 #9 · 80.90%
See charts 4 benchmarksUpdated 21 Oct 2026Reasoning & knowledgeExpert exams, graduate science, contest maths and broad knowledge.
  • Humanity's Last Exam kimi-k2.5 #17 · 24.37%
  • GPQA Diamond Kimi K3 #12 · 92.93%
  • AIME Kimi K2.5 #11 · 95.63%
  • MMLU Pro Kimi K3 #22 · 87.97%
See charts 1 benchmarkUpdated 20 Oct 2026Chat preferenceWhich answers real people prefer in blind side-by-side votes.
  • LMArena Text Arena Kimi K3 (Max) #18 · 1488
See charts

ChatJudge does not run these tests. Scores are copied as published by each source, with a link and the date we read them, and only change when we refresh from the source. Different leaderboards use different settings, so compare models within one chart, not across charts. Spotted a mistake? Tell us.