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 44.2 #20 of 339
Knowledge 5
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 43% #40 of 323+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 24% #16 of 323
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 29.3 #14 of 323
SimpleQA Verifiedv1 Accuracy on short factual questions with verified answers, answered without web access 48.9% #29 of 84
Mystery Game Puzzles Models choose the best move in puzzle positions from an undisclosed game Epoch 34% #20 of 65
Reasoning 6
ARC-AGI-1public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 96.3% #21 of 64+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 87% #33 of 67
ARC-AGI-2public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 68.3% #20 of 66+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 67.1% #23 of 69
ARC-AGI-3semi-private Measures interactive game solving through relative human action efficiency ARC 2.1% #9 of 19
Chess Puzzles Models choose the best next move from a novel chess position Epoch 33% #41 of 69
Math 3
AA Analyst Agent Data-analyst tasks: the model explores data with tools and reports findings AA 41.3% #19 of 43
FrontierMath Tier 4 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 31.7% #26 of 55
FrontierMath Tiers 1–3 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 66% #29 of 70
Science 5
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 44.1% #26 of 339
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 19.7% #32 of 326
Vals Science Index Scientific problem solving across the index’s component evaluations Vals 38.1% #10 of 19
SciCode Writing code for real scientific computing problems (physics, biology, math) AA 53% #40 of 142
Terminal-Bench Science0.1 Agents complete expert-authored research workflows across five scientific domains in a terminal AA 6.2% #23 of 45
Coding 6
Terminal-Bench4.0 Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite 17.2% #32 of 134+1 versions & slices 2.1 Older, easier version of Terminal-Bench; Still widely reported AA 88% #6 of 156
DeepSWE1.1 Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition 66.7% #17 of 44
WebDev Arena Human pairwise preferences in the stated webdev vote pool Arena 1,617 #22 of 97+2 versions & slices October 23, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,617 #23 of 98October 19, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,617 #22 of 97
Agents 7
AutomationBench-AA Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation AA 67% #9 of 135
AA-Briefcase1.1 · Elo Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 1,543 #15 of 142
EnterpriseOps-Gym-AA Agents complete stateful enterprise workflows in eight domains; final database state determines success AA 48.3% #3 of 43
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 1,647 #14 of 177
EBR Bench Agents learn to play Earthborne Rangers across repeated playthroughs Epoch 30.5% #13 of 24
Furniture Assembly Models inspect furniture assembly photos, identify incorrect builds and locate the mistaken step Epoch 40% #17 of 31
τ³-BankingAA Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 43.3% #13 of 134
Multimodal 2
Chartography Expert chart reasoning over professional scientific, medical, financial and engineering charts Surge 20.5% #21 of 37
Arena Vision Human pairwise preferences in the stated vision vote pool Arena 1,263 #37 of 88
Writing 18
Arena Textcreative writing · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,428 #41 of 197+16 versions & slices creative writing Human pairwise preferences in the stated text vote pool Arena 1,442 #35 of 195overall Human pairwise preferences in the stated text vote pool Arena 1,454 #47 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,475 #62 of 198English Human pairwise preferences in the stated text vote pool Arena 1,434 #73 of 205French Human pairwise preferences in the stated text vote pool Arena 1,452 #58 of 164German Human pairwise preferences in the stated text vote pool Arena 1,424 #58 of 167instruction following Human pairwise preferences in the stated text vote pool Arena 1,432 #53 of 205Japanese Human pairwise preferences in the stated text vote pool Arena 1,376 #68 of 152Korean Human pairwise preferences in the stated text vote pool Arena 1,385 #55 of 156non-English Human pairwise preferences in the stated text vote pool Arena 1,420 #62 of 205Polish Human pairwise preferences in the stated text vote pool Arena 1,431 #52 of 142Russian Human pairwise preferences in the stated text vote pool Arena 1,421 #65 of 203Spanish Human pairwise preferences in the stated text vote pool Arena 1,415 #77 of 168overall · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,428 #67 of 188creative writing · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,427 #43 of 187overall · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,430 #68 of 198
Lech Mazur Short-Story Judges compare stories written to ten required elements; the comparison score can be negative Lech -2.9 #41 of 45
Health & law 5
MLCR-AA Models synthesize long medical records; the headline credits concise answers judged both accurate and complete AA 17.2% #31 of 103
Vals LegalBench Legal reasoning and legal-language understanding on LegalBench Vals 86.3% #11 of 89
Vals MedCode Medical coding accuracy on patient encounters Vals 44.7% #32 of 78
Vals MedScribe Accuracy of clinical notes produced from consultation transcripts Vals 86.5% #20 of 80
Vals Legal Research Bench Grounded legal research; a task passes only if all criteria pass Vals 48.1% #8 of 73
Documents 3
AA-LCR (long context)1.1 Questions that need reading and connecting facts across very long documents (around 100K tokens) AA 81% #36 of 317
GDP.pdfAA all-pass Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes 17.2% #34 of 135
Arena Document Human pairwise preferences in the stated document vote pool Arena 1,452 #19 of 33
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 93.5% #8 of 315+1 versions & slices Epoch Models answer graduate-level science multiple-choice questions Epoch 93.2% #15 of 134
OTIS Mock AIME 20242025 Models solve the stated OTIS mock AIME mathematics problems Epoch 99.2% #14 of 131
No benchmark matches
CheaperGPT-6.1 Solmax×2.4-59%$0.72Remove unsupported parameters+5
BetterDeepSeek V4.1 Flash 1017maxOpen+44%-85%$0.26Pass back reasoning_content · Chat Completions+4
CheaperGemini 4 Argonhigh+88%-8%*$1.62Synchronous output stops at 65,536+2
GLM 5.3maxOpen-3%+14%$2.01No image input+4
Kimi K3maxOpen-24%+14%$2.00Sampling parameters are fixed+3
Qwen3.6 35B A3BnoneOpen-35%-81%$0.25 / $1.49Context: 500K → 262K