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 24.9 #85 of 337
Knowledge 5
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 34.8% #61 of 321+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 25.8% #22 of 321
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 18 #28 of 321
IFBench (instructions) Checks whether the model follows precise formatting and content instructions AA 81.3% #3 of 248
SimpleQA Verifiedv1 Accuracy on short factual questions with verified answers, answered without web access 33.2% #53 of 83
Reasoning 1
Chess Puzzles Models choose the best next move from a novel chess position Epoch 18% #56 of 68
Math 3
AA Analyst Agent Data-analyst tasks: the model explores data with tools and reports findings AA 8.8% #40 of 43
FrontierMath Tier 4 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 14.6% #46 of 54
FrontierMath Tiers 1–3 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 42.8% #50 of 69
Science 3
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 37.2% #56 of 337
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 8% #69 of 324
SciCode Writing code for real scientific computing problems (physics, biology, math) AA 48.3% #57 of 140
Coding 6
Terminal-Bench4.0 Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite 0% #75 of 132+1 versions & slices 2.1 Older, easier version of Terminal-Bench; Still widely reported AA 39.7% #85 of 156
ITBench SRE IT incident tasks: diagnose and fix broken systems like a site-reliability engineer AA 32.7% #27 of 37
WebDev Arena Human pairwise preferences in the stated webdev vote pool Arena 1,358 #79 of 97+1 versions & slices full cohort Human pairwise preferences in the stated webdev vote pool Arena 1,358 #80 of 97
Terminal-Bench Hard Agents solve difficult software, system-administration and data-processing tasks in a terminal AA 37.9% #36 of 242
Agents 6
AutomationBench-AA Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation AA 1.4% #106 of 133
AA-Briefcase1.1 · Elo Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 762 #78 of 140
APEX-Agents Agents complete long-horizon professional-services tasks across realistic applications AA 17% #19 of 29
EnterpriseOps-Gym-AA Agents complete stateful enterprise workflows in eight domains; final database state determines success AA 26.3% #40 of 43
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 21.9 #89 of 175
τ³-BankingAA Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 12.4% #81 of 134
Multimodal 3
MMMU-Pro (vision) College-level questions that require understanding images: charts, diagrams, photos AA 78.1% #34 of 127
Arena Vision Human pairwise preferences in the stated vision vote pool Arena 1,229 #55 of 88
Chartography Expert chart reasoning over professional scientific, medical, financial and engineering charts Surge 12.9% #32 of 37
Writing 16
Arena Textcreative writing Human pairwise preferences in the stated text vote pool Arena 1,422 #52 of 195+14 versions & slices overall Human pairwise preferences in the stated text vote pool Arena 1,442 #61 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,421 #96 of 181English Human pairwise preferences in the stated text vote pool Arena 1,404 #95 of 188French Human pairwise preferences in the stated text vote pool Arena 1,412 #83 of 149German Human pairwise preferences in the stated text vote pool Arena 1,395 #74 of 154instruction following Human pairwise preferences in the stated text vote pool Arena 1,366 #99 of 188Japanese Human pairwise preferences in the stated text vote pool Arena 1,379 #60 of 138Korean Human pairwise preferences in the stated text vote pool Arena 1,355 #76 of 144non-English Human pairwise preferences in the stated text vote pool Arena 1,385 #90 of 188Polish Human pairwise preferences in the stated text vote pool Arena 1,398 #73 of 133Russian Human pairwise preferences in the stated text vote pool Arena 1,399 #83 of 186Spanish Human pairwise preferences in the stated text vote pool Arena 1,399 #88 of 155overall · full cohort Human pairwise preferences in the stated text vote pool Arena 1,397 #93 of 188creative writing · full cohort Human pairwise preferences in the stated text vote pool Arena 1,380 #78 of 187
Lech Mazur Short-Story Judges compare stories written to ten required elements; the comparison score can be negative Lech -4.3 #43 of 44
Health & law 5
MLCR-AA Models synthesize long medical records; the headline credits concise answers judged both accurate and complete AA 10% #41 of 64
Vals LegalBench Legal reasoning and legal-language understanding on LegalBench Vals 84.5% #28 of 89
Vals MedCode Medical coding accuracy on patient encounters Vals 38.1% #56 of 78
Vals MedScribe Accuracy of clinical notes produced from consultation transcripts Vals 74.4% #62 of 80
Vals Legal Research Bench Grounded legal research; a task passes only if all criteria pass Vals 15.4% #54 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 73% #101 of 275
GDP.pdfAA all-pass Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes 5.8% #80 of 133
Legacy · saturated or no longer reported 4
GPQA Diamond Graduate-level biology, chemistry and physics questions that can’t be solved by a quick web search 90.1% #34 of 315+1 versions & slices Epoch Models answer graduate-level science multiple-choice questions Epoch 88.8% #38 of 133
OTIS Mock AIME 20242025 Models solve the stated OTIS mock AIME mathematics problems Epoch 93.3% #38 of 130
τ²-benchAA Tests service agents that use tools while coordinating with a simulated user AA 97.7% #9 of 245
No benchmark matches
CheaperGPT-5.6 Lunamax+92%-13%$0.18
OpenKimi K3max+69%+874%$2.00Sampling parameters are fixed+3
CheaperGemini 3 Flash Previewunstated+20%-28%$0.50 / $3
Claude Haiku 5.5max+75%+4%$0.21Remove sampling parameters+5
GLM 5.3 FlashmaxOpen+30%+38%$0.28Defaults to max · Self-hosted+4
DeepSeek V4 Pro 0813maxOpen+51%+228%$0.67No image input+5