← Benchmarks
Report cards
Benchmark report cards
Every axis the console can rank on, with the facts behind it: who runs it, whether the test set is exposed, when it last moved, how many models it covers, whether it publishes intervals, and which other axes it duplicates. No score about the benchmark — the ledger is the verdict.
Overall
- Overall
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 416
- Newest result
- 4 days ago
- Intervals
- on every row
Ability
- Coding
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 411
- Newest result
- 4 days ago
- Intervals
- on every row
- Instruction following
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 416
- Newest result
- 4 days ago
- Intervals
- on every row
- Creative writing
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 414
- Newest result
- 4 days ago
- Intervals
- on every row
- Hard prompts
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 416
- Newest result
- 4 days ago
- Intervals
- on every row
- Math
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 400
- Newest result
- 4 days ago
- Intervals
- on every row
- Multi-turn
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 414
- Newest result
- 4 days ago
- Intervals
- on every row
- Expert
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 366
- Newest result
- 4 days ago
- Intervals
- on every row
Industry
- Software & IT
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 416
- Newest result
- 4 days ago
- Intervals
- on every row
- Business & finance
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 409
- Newest result
- 4 days ago
- Intervals
- on every row
- Writing & language
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 415
- Newest result
- 4 days ago
- Intervals
- on every row
- Entertainment & media
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 414
- Newest result
- 4 days ago
- Intervals
- on every row
Language
- English
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 416
- Newest result
- 4 days ago
- Intervals
- on every row
- Chinese
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 392
- Newest result
- 4 days ago
- Intervals
- on every row
- Spanish
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 301
- Newest result
- 4 days ago
- Intervals
- on every row
- French
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 297
- Newest result
- 4 days ago
- Intervals
- on every row
- German
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 314
- Newest result
- 4 days ago
- Intervals
- on every row
- Japanese
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 280
- Newest result
- 4 days ago
- Intervals
- on every row
- Korean
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 284
- Newest result
- 4 days ago
- Intervals
- on every row
- Russian
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 380
- Newest result
- 4 days ago
- Intervals
- on every row
Arena boards
- WebDev Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 284
- Newest result
- 3 days ago
- Intervals
- on every row
- Vision Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 165
- Newest result
- Oct 9, 2026
- Intervals
- on every row
- Factuality Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 185
- Newest result
- 4 days ago
- Intervals
- on every row
- Search Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 42
- Newest result
- 6 days ago
- Intervals
- on every row
- Document Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 44
- Newest result
- Sep 13, 2026
- Intervals
- on every row
- Text-to-Image Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 82
- Newest result
- Oct 7, 2026
- Intervals
- on every row
- Image Edit Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 59
- Newest result
- Oct 6, 2026
- Intervals
- on every row
- Text-to-Video Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 48
- Newest result
- Sep 22, 2026
- Intervals
- on every row
Measured benchmarks
- GPQA Diamond
Epoch AI
- Run by
- an independent organisation
- Test set
- public
- Models
- 324
- Newest result
- 3 days ago
- Intervals
- on every row
- AIME (math)
Epoch AI
- Run by
- an independent organisation
- Test set
- public
- Models
- 304
- Newest result
- 3 days ago
- Intervals
- 95.7% of rows
- FrontierMath
Epoch AI
- Run by
- an independent organisation
- Test set
- held out
- Models
- 101
- Newest result
- Jun 8, 2026
- Intervals
- 98% of rows
- ARC-AGI-2
ARC Prize
- Run by
- an independent organisation
- Test set
- part public, part held out
- Models
- 267
- Newest result
- Sep 29, 2026
- Intervals
- none published
- Humanity's Last Exam
Scale AI / CAIS
- Run by
- an independent organisation
- Test set
- part public, part held out
- Models
- 56
- Newest result
- Oct 7, 2026
- Intervals
- on every row
- SimpleBench
SimpleBench
- Run by
- an independent organisation
- Test set
- held out
- Models
- 102
- Newest result
- Aug 12, 2026
- Intervals
- none published
- SimpleQA Verified
Epoch AI
- Run by
- an independent organisation
- Test set
- public
- Models
- 91
- Newest result
- 3 days ago
- Intervals
- on every row
- MMLU-Pro (TIGER-Lab)
TIGER-Lab (University of Waterloo)
- Run by
- a university group
- Test set
- public
- Models
- 260
- Newest result
- not published
- Intervals
- none published
- LiveBench
LiveBench
- Run by
- a university group
- Test set
- live / refreshed
- Models
- 339
- Newest result
- Sep 27, 2026
- Intervals
- none published
- MMLU-Pro (Artificial Analysis)
Artificial Analysis
- Run by
- an independent organisation
- Test set
- public
- Models
- 352
- Newest result
- 2 days ago
- Intervals
- none published
- LiveCodeBench (Artificial Analysis)
Artificial Analysis
- Run by
- an independent organisation
- Test set
- live / refreshed
- Models
- 346
- Newest result
- 2 days ago
- Intervals
- none published
- IFBench (Artificial Analysis)
Artificial Analysis
- Run by
- an independent organisation
- Test set
- public
- Models
- 455
- Newest result
- 2 days ago
- Intervals
- none published
- USACO (HAL)
Holistic Agent Leaderboard (Princeton)
- Run by
- a university group
- Test set
- public
- Models
- 9
- Newest result
- Sep 27, 2026
- Intervals
- none published
- Output speed, tokens/s (Artificial Analysis)
Artificial Analysis
- Run by
- an independent organisation
- Test set
- live / refreshed
- Models
- 467
- Newest result
- 2 days ago
- Intervals
- none published
Agentic
- Agent Arena
arena.ai (LMArena)
- Run by
- public votes
- Test set
- live / refreshed
- Models
- 80
- Newest result
- 3 days ago
- Intervals
- on every row
- SWE-bench Verified
Epoch AI
- Run by
- an independent organisation
- Test set
- public
- Models
- 35
- Newest result
- Jun 25, 2026
- Intervals
- on every row
- Terminal-Bench
Terminal-Bench (Laude Institute)
- Run by
- an independent organisation
- Test set
- public
- Models
- 202
- Newest result
- May 14, 2026
- Intervals
- 95.5% of rows
- SWE-bench Verified (live board)
SWE-bench Verified leaderboard
- Run by
- a university group
- Test set
- public
- Models
- 85
- Newest result
- Sep 1, 2026
- Intervals
- none published
- Terminal-Bench 4.0
Terminal-Bench 4.0 (Harbor / Laude Institute)
- Run by
- an independent organisation
- Test set
- public
- Models
- 23
- Newest result
- Sep 29, 2026
- Intervals
- on every row
- BFCL v4 (function calling)
Berkeley Function Calling Leaderboard (Gorilla, UC Berkeley)
- Run by
- a university group
- Test set
- part public, part held out
- Models
- 109
- Newest result
- Apr 3, 2026dormant
- Intervals
- none published
- τ²-bench Retail
τ²-bench via the Exgentic Open Agent Leaderboard
- Run by
- a model vendor
- Test set
- public
- Models
- 3
- Newest result
- Mar 23, 2026dormant
- Intervals
- none published
- τ²-bench Airline
τ²-bench via the Exgentic Open Agent Leaderboard
- Run by
- a model vendor
- Test set
- public
- Models
- 3
- Newest result
- Mar 23, 2026dormant
- Intervals
- none published
- τ²-bench Telecom
τ²-bench via the Exgentic Open Agent Leaderboard
- Run by
- a model vendor
- Test set
- public
- Models
- 3
- Newest result
- Mar 23, 2026dormant
- Intervals
- none published
- GAIA (HAL)
Holistic Agent Leaderboard (Princeton)
- Run by
- a university group
- Test set
- part public, part held out
- Models
- 12
- Newest result
- Sep 27, 2026
- Intervals
- none published
Safety & factuality
- HHEM Hallucination (Vectara)
Vectara
- Run by
- a model vendor
- Test set
- public
- Models
- 109
- Newest result
- 3 days ago
- Intervals
- none published