Benchmark By Score Rank
Overall 1
AA Intelligence Index4.3.2 Artificial Analysis combines tests of knowledge, reasoning, coding and tool use into one score AA 8.4 #239 of 337
Knowledge 5
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 19.9% #157 of 321+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 37.9% #57 of 321
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA -10.5 #95 of 321
IFBench (instructions) Checks whether the model follows precise formatting and content instructions AA 34.3% #191 of 248
SimpleQA Verifiedv1 Accuracy on short factual questions with verified answers, answered without web access 26% #63 of 83
Reasoning 4
ARC-AGI-1semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 4.5% #61 of 64
ARC-AGI-2public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 0% #54 of 63+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 0% #59 of 66
Chess Puzzles Models choose the best next move from a novel chess position Epoch 6% #63 of 68
Math 3
FrontierMath Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 0.3% #39 of 40
FrontierMath Tiers 1–3 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 0.4% #69 of 69
MATH Level5 Models solve competition mathematics problems from the hardest MATH difficulty level Epoch 53.3% #19 of 35
Science 2
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 2.4% #335 of 337
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 0% #180 of 324
Coding 2
SWE-bench Pro (private) Agents resolve long-horizon professional repository tasks; public and private splits are separate Scale 3.6% #9 of 9
Terminal-Bench Hard Agents solve difficult software, system-administration and data-processing tasks in a terminal AA 8.3% #133 of 242
Agents 1
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 0 #122 of 175
Multimodal 4
Arena Vision Human pairwise preferences in the stated vision vote pool Arena 1,137 #76 of 88
MMMUvalidation Image-grounded university-level questions across 30 disciplines; validation and test sets remain separate MMMU 70.7% #13 of 41
OCRBenchv2 · CN Tests visual text recognition, localization and reasoning; English and Chinese subsets remain separate OCRBench 32.2% #3 of 4+1 versions & slices v2 · EN Tests visual text recognition, localization and reasoning; English and Chinese subsets remain separate OCRBench 46.5% #2 of 4
Writing 15
Arena Textcreative writing Human pairwise preferences in the stated text vote pool Arena 1,338 #106 of 195+14 versions & slices overall Human pairwise preferences in the stated text vote pool Arena 1,346 #126 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,277 #141 of 181English Human pairwise preferences in the stated text vote pool Arena 1,312 #144 of 188French Human pairwise preferences in the stated text vote pool Arena 1,304 #118 of 149German Human pairwise preferences in the stated text vote pool Arena 1,282 #115 of 154instruction following Human pairwise preferences in the stated text vote pool Arena 1,278 #138 of 188Japanese Human pairwise preferences in the stated text vote pool Arena 1,257 #99 of 138Korean Human pairwise preferences in the stated text vote pool Arena 1,234 #114 of 144non-English Human pairwise preferences in the stated text vote pool Arena 1,283 #138 of 188Polish Human pairwise preferences in the stated text vote pool Arena 1,365 #88 of 133Russian Human pairwise preferences in the stated text vote pool Arena 1,286 #137 of 186Spanish Human pairwise preferences in the stated text vote pool Arena 1,292 #127 of 155overall · full cohort Human pairwise preferences in the stated text vote pool Arena 1,300 #139 of 188creative writing · full cohort Human pairwise preferences in the stated text vote pool Arena 1,291 #125 of 187
Health & law 1
Vals LegalBench Legal reasoning and legal-language understanding on LegalBench Vals 82.2% #53 of 89
Documents 1
AA-LCR (long context)1.1 Questions that need reading and connecting facts across very long documents (around 100K tokens) AA 49.3% #181 of 275
Legacy · saturated or no longer reported 12
AIME2025 · AA archived AA evaluates the 30 integer-answer problems from the 2025 AIME mathematics competition AA 6% #131 of 140
LiveCodeBenchAA archived AA evaluates solutions to newly collected competitive-programming problems AA 30.9% #118 of 153+1 versions & slices official artifact (through April 2025) Mean pass@1 over each model’s available problems in the current official public generation artifact; its latest problem is dated April 7, 2025 LiveCodeBench 38.3% #9 of 10
GPQA Diamond Graduate-level biology, chemistry and physics questions that can’t be solved by a quick web search 54.3% #247 of 315+1 versions & slices Epoch Models answer graduate-level science multiple-choice questions Epoch 49.2% #103 of 133
Aider Polyglottwo attempts Models edit code for 225 Exercism exercises in six languages, with a second attempt after test feedback Aider 23.1% #10 of 15
MMLU-Proofficial board Graduate-level multiple-choice knowledge and reasoning across 14 subjects, including science, mathematics, law and health 74.7% #2 of 32
OTIS Mock AIME 20242025 Models solve the stated OTIS mock AIME mathematics problems Epoch 6.4% #113 of 130
SWE-bench VerifiedEpoch Swe Bench Verified: Agents modify real repositories to resolve the stated Verified issues Epoch 31% #24 of 24
MathArenaAIME 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 11.7% #30 of 30+1 versions & slices HMMT February 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 5.8% #30 of 30
τ²-benchAA Tests service agents that use tools while coordinating with a simulated user AA 25.1% #174 of 245
No benchmark matches
CheaperQwen3 VL 235B A22B+11%-76%$0.53 / $2.66
CheaperLlama 4 MaverickunstatedOpen+4%-93%$0.19 / $0.65
Claude Opus 4.5unstated+14%+129%$5 / $25dDots.vlm1.instOpen+13%——
Gemini 2.5 Pro Preview (May' 25)unstated+13%——No image input
Skywork R1V3 38BOpen+7%——Context: 128K → 33K