Benchmark By Score Rank
Overall 1
AA Intelligence Index4.3.2 AA Intelligence Index 4.3.2: Artificial Analysis combines tests of knowledge, reasoning, coding and tool use into one score AA 42.1 #24 of 337
Knowledge 6
AA OmniscienceAccuracy AA Omniscience · Accuracy: Measures correct answers on the AA Omniscience factual knowledge test AA 46.8% #31 of 321+1 versions & slices Hallucination Rate AA Omniscience · Hallucination Rate: Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 87.9% #228 of 321
AA Omniscience Index AA Omniscience Index: Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 0.1 #72 of 321
IFBench (instructions) IFBench (instructions): Checks whether the model follows precise formatting and content instructions AA 71.2% #44 of 248
SimpleQA Verifiedv1 SimpleQA Verified v1: Accuracy on short factual questions with verified answers, answered without web access 43.2% #37 of 83
Mystery Game Puzzles Mystery Game Puzzles: Models choose the best move in puzzle positions from an undisclosed game Epoch 35% #17 of 64
Reasoning 6
ARC-AGI-1public evaluation ARC-AGI-1 · public-eval: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 98.3% #8 of 61+1 versions & slices semi-private ARC-AGI-1 · semi-private: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 96.5% #11 of 64
ARC-AGI-2public evaluation ARC-AGI-2 · public-eval: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 91.3% #11 of 63+1 versions & slices semi-private ARC-AGI-2 · semi-private: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 83.9% #14 of 66
ARC-AGI-3semi-private ARC-AGI-3 · semi-private: Measures interactive game solving through relative human action efficiency ARC 0.8% #11 of 18
Chess Puzzles Chess Puzzles: Models choose the best next move from a novel chess position Epoch 47% #19 of 68
Math 2
FrontierMath Tier 4 (v2) FrontierMath Tier 4 (v2): Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 70.7% #13 of 54
FrontierMath Tiers 1–3 (v2) FrontierMath Tiers 1–3 (v2): Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 86% #10 of 69
Science 4
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models AA 42.9% #32 of 337
CritPt (physics) CritPt (physics): Research-level physics problems that need many careful reasoning steps AA 30% #10 of 324
SciCode SciCode: Writing code for real scientific computing problems (physics, biology, math) AA 55% #27 of 140
Terminal-Bench Science0.1 Terminal-Bench Science 0.1: Agents complete expert-authored research workflows across five scientific domains in a terminal AA 13.8% #11 of 33
Coding 7
Terminal-Bench4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite 35.4% #15 of 132+1 versions & slices 2.1 Terminal-Bench 2.1 · source setting 2: Older, easier version of Terminal-Bench; Still widely reported AA 88% #7 of 156
DeepSWE1.1 DeepSWE 1.1: Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition 69.6% #10 of 44
ITBench SRE ITBench SRE: IT incident tasks: diagnose and fix broken systems like a site-reliability engineer AA 51% #5 of 37
SWE-bench Pro (v2 full) SWE-bench Pro (v2 full): Agents resolve repository issues under the network-locked v2 protocol; full and 51-task Hard results are separate Scale 92.4% #14 of 16
SWE-bench Pro (v2 hard) SWE-bench Pro (v2 hard): Agents resolve repository issues under the network-locked v2 protocol; full and 51-task Hard results are separate Scale 86.3% #8 of 16
Terminal-Bench Hard Terminal-Bench Hard (version unstated): Agents solve difficult software, system-administration and data-processing tasks in a terminal AA 57.6% #5 of 242
Agents 7
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation AA 59.6% #20 of 133
AA-Briefcase1.1 · Elo AA-Briefcase 1.1 · Elo: Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 1,334 #34 of 140
APEX-Agents APEX-Agents: Agents complete long-horizon professional-services tasks across realistic applications AA 38.9% #3 of 29
EnterpriseOps-Gym-AA EnterpriseOps-Gym-AA: Agents complete stateful enterprise workflows in eight domains; final database state determines success AA 38.5% #25 of 43
GDPval-AA (work tasks)2.1 GDPval-AA (work tasks) 2.1: Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 47.6 #33 of 175
Furniture Assembly Furniture Assembly: Models inspect furniture assembly photos, identify incorrect builds and locate the mistaken step Epoch 54.2% #10 of 31
τ³-BankingAA τ³-Banking · AA: Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 40.2% #16 of 134
Multimodal 3
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos AA 80.7% #14 of 127
Arena Vision Arena Vision · overall: Human pairwise preferences in the stated vision vote pool Arena 1,271 #30 of 88
Chartography Chartography: Expert chart reasoning over professional scientific, medical, financial and engineering charts Surge 34% #9 of 37
Writing 17
Arena Textcreative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool Arena 1,429 #46 of 195+14 versions & slices overall Arena Text · overall: Human pairwise preferences in the stated text vote pool Arena 1,466 #34 of 196Chinese Arena Text · chinese: Human pairwise preferences in the stated text vote pool Arena 1,515 #25 of 181English Arena Text · english: Human pairwise preferences in the stated text vote pool Arena 1,448 #51 of 188French Arena Text · french: Human pairwise preferences in the stated text vote pool Arena 1,469 #37 of 149German Arena Text · german: Human pairwise preferences in the stated text vote pool Arena 1,458 #29 of 154instruction following Arena Text · instruction following: Human pairwise preferences in the stated text vote pool Arena 1,456 #33 of 188Japanese Arena Text · japanese: Human pairwise preferences in the stated text vote pool Arena 1,456 #16 of 138Korean Arena Text · korean: Human pairwise preferences in the stated text vote pool Arena 1,427 #25 of 144non-English Arena Text · non english: Human pairwise preferences in the stated text vote pool Arena 1,440 #35 of 188Polish Arena Text · polish: Human pairwise preferences in the stated text vote pool Arena 1,447 #31 of 133Russian Arena Text · russian: Human pairwise preferences in the stated text vote pool Arena 1,452 #34 of 186Spanish Arena Text · spanish: Human pairwise preferences in the stated text vote pool Arena 1,450 #44 of 155overall · full cohort Arena Text · overall · snapshot 2026-10-15 · native board settings: Human pairwise preferences in the stated text vote pool Arena 1,447 #40 of 188creative writing · full cohort Arena Text · creative writing · snapshot 2026-10-15 · native board settings: Human pairwise preferences in the stated text vote pool Arena 1,412 #50 of 187
EQ-Bench Longform Creative Writing1.11 Longform Creative Writing 1.11: Models plan, revise and write a story across eight turns, graded with a writing rubric EQ-Bench 78% #21 of 95
EQ-Bench Creative Writingv3 Creative Writing v3 · Elo: Judge-rated creative writing under the v3 prompt set and pairwise ranking EQ-Bench 1,855 #16 of 91
Health & law 5
MLCR-AA MLCR-AA: Models synthesize long medical records; the headline credits concise answers judged both accurate and complete AA 31.7% #15 of 64
Vals LegalBench Vals LegalBench: Legal reasoning and legal-language understanding on LegalBench Vals 85.1% #21 of 89
Vals MedCode Vals MedCode: Medical coding accuracy on patient encounters Vals 43.4% #37 of 78
Vals MedScribe Vals MedScribe: Accuracy of clinical notes produced from consultation transcripts Vals 82.9% #43 of 80
Vals Legal Research Bench Vals Legal Research Bench: Grounded legal research; a task passes only if all criteria pass Vals 41.4% #22 of 69
Documents 3
AA-LCR (long context)1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens) AA 83% #17 of 275
GDP.pdfAA all-pass GDP.pdf · AA all-pass: Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes 24% #11 of 133
Arena Document Arena Document · overall: Human pairwise preferences in the stated document vote pool Arena 1,472 #10 of 33
Legacy · saturated or no longer reported 4
GPQA Diamond GPQA Diamond: Graduate-level biology, chemistry and physics questions that can’t be solved by a quick web search 92.5% #18 of 315+1 versions & slices Epoch GPQA Diamond · source setting 2: Models answer graduate-level science multiple-choice questions Epoch 93.3% #14 of 133
OTIS Mock AIME 20242025 OTIS Mock AIME 2024 2025: Models solve the stated OTIS mock AIME mathematics problems Epoch 99.7% #11 of 130
τ²-benchAA τ²-Bench: Tests service agents that use tools while coordinating with a simulated user AA 86.3% #59 of 245
No benchmark matches
BetterMiMo v2.6 ProthinkingOpen+10%-90%$0.13Context: 1.05M → 1M+4
CheaperClaude Sonnet 5.5high+11%-37%$0.88Context: 1.05M → 1M+6
DeepSeek V4.1 Flash 1017maxOpen+7%-81%$0.26Context: 1.05M → 1M+5
GLM 5.3maxOpen+6%+43%$2.01Context: 1.05M → 1M+5
Step 5 PreviewthinkingOpen+4%-27%$1.03Context: 1.05M → 1M+4
Kimi K3maxOpen+4%+43%$2.00Context: 1.05M → 1M+4