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 22.2 #103 of 337
Knowledge 5
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 29.5% #81 of 321+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 34.4% #48 of 321
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 5.2 #51 of 321
SimpleQA Verifiedv1 Accuracy on short factual questions with verified answers, answered without web access 33.1% #54 of 83
Mystery Game Puzzles Models choose the best move in puzzle positions from an undisclosed game Epoch 19% #38 of 64
Reasoning 5
ARC-AGI-1public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 66.4% #42 of 61+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 53.5% #49 of 64
ARC-AGI-2public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 7.9% #43 of 63+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 10.3% #45 of 66
Chess Puzzles Models choose the best next move from a novel chess position Epoch 16% #56 of 68
Math 2
FrontierMath Tier 4 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 0% #53 of 54
FrontierMath Tiers 1–3 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 26% #59 of 69
Science 4
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 18.8% #128 of 337
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 0% #180 of 324
SciCode Writing code for real scientific computing problems (physics, biology, math) AA 41.3% #90 of 140
Terminal-Bench Science0.1 Agents complete expert-authored research workflows across five scientific domains in a terminal AA 0% #27 of 33
Coding 4
Terminal-Bench4.0 Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite 1% #54 of 132+1 versions & slices 2.1 Older, easier version of Terminal-Bench; Still widely reported AA 53.6% #65 of 156
WebDev Arena Human pairwise preferences in the stated webdev vote pool Arena 1,435 #58 of 97+1 versions & slices full cohort Human pairwise preferences in the stated webdev vote pool Arena 1,435 #60 of 97
Agents 5
AutomationBench-AA Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation AA 25% #53 of 133
AA-Briefcase1.1 · Elo Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 650 #88 of 140
EnterpriseOps-Gym-AA Agents complete stateful enterprise workflows in eight domains; final database state determines success AA 42.3% #18 of 43
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 24.2 #83 of 175
τ³-BankingAA Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 17.5% #61 of 134
Multimodal 3
MMMU-Pro (vision) College-level questions that require understanding images: charts, diagrams, photos AA 79% #26 of 127
Arena Vision Human pairwise preferences in the stated vision vote pool Arena 1,269 #32 of 88
Chartography Expert chart reasoning over professional scientific, medical, financial and engineering charts Surge 10.2% #35 of 37
Writing 16
Arena Textcreative writing Human pairwise preferences in the stated text vote pool Arena 1,436 #40 of 195+14 versions & slices overall Human pairwise preferences in the stated text vote pool Arena 1,455 #46 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,467 #66 of 181English Human pairwise preferences in the stated text vote pool Arena 1,438 #67 of 188French Human pairwise preferences in the stated text vote pool Arena 1,451 #55 of 149German Human pairwise preferences in the stated text vote pool Arena 1,454 #33 of 154instruction following Human pairwise preferences in the stated text vote pool Arena 1,419 #65 of 188Japanese Human pairwise preferences in the stated text vote pool Arena 1,427 #29 of 138Korean Human pairwise preferences in the stated text vote pool Arena 1,407 #37 of 144non-English Human pairwise preferences in the stated text vote pool Arena 1,426 #53 of 188Polish Human pairwise preferences in the stated text vote pool Arena 1,437 #40 of 133Russian Human pairwise preferences in the stated text vote pool Arena 1,443 #42 of 186Spanish Human pairwise preferences in the stated text vote pool Arena 1,441 #52 of 155overall · full cohort Human pairwise preferences in the stated text vote pool Arena 1,435 #59 of 188creative writing · full cohort Human pairwise preferences in the stated text vote pool Arena 1,420 #45 of 187
EQ-Bench Creative Writingv3 Creative Writing v3 · Elo: Judge-rated creative writing under the v3 prompt set and pairwise ranking EQ-Bench 1,559 #46 of 91
Health & law 6
Harvey LAB-AA1.1 Legal agents work from case documents; every rubric criterion must pass and material hallucinations receive no credit AA 0% #18 of 23
MLCR-AA Models synthesize long medical records; the headline credits concise answers judged both accurate and complete AA 7.2% #45 of 64
Vals LegalBench Legal reasoning and legal-language understanding on LegalBench Vals 84.1% #35 of 89
Vals MedCode Medical coding accuracy on patient encounters Vals 43.5% #36 of 78
Vals MedScribe Accuracy of clinical notes produced from consultation transcripts Vals 70.9% #68 of 80
Vals Legal Research Bench Grounded legal research; a task passes only if all criteria pass Vals 13.9% #58 of 69
Documents 2
AA-LCR (long context)1.1 Questions that need reading and connecting facts across very long documents (around 100K tokens) AA 76% #84 of 275
GDP.pdfAA all-pass Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes 13.6% #42 of 133
Legacy · saturated or no longer reported 3
GPQA Diamond Graduate-level biology, chemistry and physics questions that can’t be solved by a quick web search 83.8% #91 of 315+1 versions & slices Epoch Models answer graduate-level science multiple-choice questions Epoch 83.3% #60 of 133
OTIS Mock AIME 20242025 Models solve the stated OTIS mock AIME mathematics problems Epoch 71.1% #74 of 130
No benchmark matches
CheaperGPT-5.6 Lunamax×3.2-7%$0.18
OpenKimi K3max×2.8+938%$2.00Sampling parameters are fixed+3
Grok 4.3high+65%+7%$0.21
Claude Haiku 5.5max×2.9+10%$0.21Remove sampling parameters+5
GLM 5.3 FlashmaxOpen×2.1+47%$0.28Defaults to max · Self-hosted+4
DeepSeek V4 Pro 0813maxOpen×2.5+250%$0.67No image input+6