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 42.1 #25 of 339
Knowledge 5
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 58.9% #7 of 323+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 51.6% #95 of 323
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 37.6 #6 of 323
SimpleQA Verifiedv1 Accuracy on short factual questions with verified answers, answered without web access 73.9% #5 of 84
Mystery Game Puzzles Models choose the best move in puzzle positions from an undisclosed game Epoch 80% #2 of 65
Reasoning 6
ARC-AGI-1public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 96.4% #20 of 64+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 93.5% #21 of 67
ARC-AGI-2public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 77.8% #19 of 66+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 76.7% #19 of 69
ARC-AGI-3semi-private Measures interactive game solving through relative human action efficiency ARC 3.9% #8 of 19
Chess Puzzles Models choose the best next move from a novel chess position Epoch 68% #2 of 69
Math 13 Leads on 4
AA Analyst Agent Data-analyst tasks: the model explores data with tools and reports findings AA 50% #8 of 43
FrontierMath Erdos Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 4.4% #1 of 13
FrontierMath Tier 4 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 100% #1 of 55
FrontierMath Tiers 1–3 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 93.7% #1 of 70
MathArenaArXivMath · August 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 96.5% #1 of 14+8 versions & slices ArXivMath · June 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 94.4% #1 of 24ArXivMath · May 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 95% #1 of 19ArXivMath · September 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 91.5% #1 of 9BrokenArXiv · August 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 86.3% #1 of 13BrokenArXiv · June 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 100% #1 of 24BrokenArXiv · May 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 90% #3 of 18BrokenArXiv · September 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 82.7% #1 of 9ArXivLean · September 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 35.9% #4 of 6
Science 5
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 47.4% #17 of 339
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 24.9% #21 of 326
Vals Science Index Scientific problem solving across the index’s component evaluations Vals 59.6% #2 of 19
SciCode Writing code for real scientific computing problems (physics, biology, math) AA 53.2% #40 of 142
Terminal-Bench Science0.1 Agents complete expert-authored research workflows across five scientific domains in a terminal AA 11% #16 of 45
Coding 7
Terminal-Bench4.0 Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite 30.8% #24 of 134
DeepSWE1.1 Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition 72.3% #5 of 44
FrontierCode1.1 Agents implement changes in real repositories, graded against maintainer-defined requirements AA 36.7% #25 of 38
WebDev Arena Human pairwise preferences in the stated webdev vote pool Arena 1,753 #4 of 97+2 versions & slices October 23, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,749 #4 of 98October 19, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,753 #4 of 97
SWE-bench Pro (v2 full) Agents resolve repository issues under the network-locked v2 protocol; full and 51-task Hard results are separate Scale 96.1% #7 of 16
Agents 8 Leads on 1
AutomationBench-AA Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation AA 52.6% #35 of 135
AA-Briefcase1.1 · Elo Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 1,114 #48 of 142
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 1,323 #46 of 177
EBR Bench Agents learn to play Earthborne Rangers across repeated playthroughs Epoch 54.3% #5 of 24
Furniture Assembly Models inspect furniture assembly photos, identify incorrect builds and locate the mistaken step Epoch 80% #2 of 31
SpireClimb Agents play Slay the Spire under the named game and difficulty setting Epoch 24.7% #2 of 7
SpireClimb StS1 A20 Heart Agents play Slay the Spire under the named game and difficulty setting Epoch 20% #1 of 7
SpireClimb StS2 A10 Agents play Slay the Spire under the named game and difficulty setting Epoch 29.3% #2 of 7
Multimodal 2
MMMU-Pro (vision) College-level questions that require understanding images: charts, diagrams, photos AA 83.1% #8 of 128
Arena Vision Human pairwise preferences in the stated vision vote pool Arena 1,288 #18 of 88
Writing 13
Arena Textcreative writing · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,424 #44 of 197+12 versions & slices creative writing Human pairwise preferences in the stated text vote pool Arena 1,458 #19 of 195overall Human pairwise preferences in the stated text vote pool Arena 1,484 #17 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,484 #53 of 198English Human pairwise preferences in the stated text vote pool Arena 1,451 #49 of 205French Human pairwise preferences in the stated text vote pool Arena 1,461 #44 of 164instruction following Human pairwise preferences in the stated text vote pool Arena 1,467 #24 of 205non-English Human pairwise preferences in the stated text vote pool Arena 1,435 #41 of 205Russian Human pairwise preferences in the stated text vote pool Arena 1,458 #30 of 203Spanish Human pairwise preferences in the stated text vote pool Arena 1,445 #53 of 168overall · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,448 #39 of 188creative writing · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,430 #41 of 187overall · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,444 #45 of 198
Health & law 5
Harvey LAB-AA1.1 Legal agents work from case documents; every rubric criterion must pass and material hallucinations receive no credit AA 6.9% #6 of 23
MLCR-AA Models synthesize long medical records; the headline credits concise answers judged both accurate and complete AA 26.7% #12 of 103
Vals MedCode Medical coding accuracy on patient encounters Vals 48.8% #24 of 78
Vals MedScribe Accuracy of clinical notes produced from consultation transcripts Vals 86.5% #21 of 80
Vals Legal Research Bench Grounded legal research; a task passes only if all criteria pass Vals 38.5% #29 of 73
Documents 2
AA-LCR (long context)1.1 Questions that need reading and connecting facts across very long documents (around 100K tokens) AA 84% #7 of 317
GDP.pdfAA all-pass Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes 27% #8 of 135
Legacy · saturated or no longer reported 2
GPQA DiamondEpoch Models answer graduate-level science multiple-choice questions Epoch 95.4% #4 of 134
OTIS Mock AIME 20242025 Models solve the stated OTIS mock AIME mathematics problems Epoch 100% #1 of 131
No benchmark matches
BetterMiMo v2.6 ProthinkingOpen+10%same$0.13Context: 1.05M → 1M+4
CheaperClaude Haiku 5.5xhighsame-5%$0.12Context: 1.05M → 1M+6
Agnes 2.5 Pro Beta-16%-96%$0.10 / $0.30Context: 1.05M → 1M
DeepSeek V4.1 Flash 1017maxOpen+7%+100%$0.26Context: 1.05M → 1M+5
GLM 5.3maxOpen+6%+1,434%$2.01Context: 1.05M → 1M+5
Step 5 PreviewthinkingOpen+4%+685%$1.03Context: 1.05M → 1M+4
Kimi K3maxOpen+4%+1,430%$2.00Context: 1.05M → 1M+4