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 29.2 #62 of 339
Knowledge 5
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 16.7% #208 of 323+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 18.4% #9 of 323
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 1.4 #63 of 323
IFBench (instructions) Checks whether the model follows precise formatting and content instructions AA 82.9% #2 of 248
Mystery Game Puzzles Models choose the best move in puzzle positions from an undisclosed game Epoch 8% #59 of 65
Reasoning 1
Chess Puzzles Models choose the best next move from a novel chess position Epoch 7% #63 of 69
Math 1
AA Analyst Agent Data-analyst tasks: the model explores data with tools and reports findings AA 10% #39 of 43
Science 4
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 39% #51 of 339
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 3.7% #90 of 326
SciCode Writing code for real scientific computing problems (physics, biology, math) AA 47.1% #67 of 142
Terminal-Bench Science0.1 Agents complete expert-authored research workflows across five scientific domains in a terminal AA 0.5% #34 of 45
Coding 9
Terminal-Bench4.0 Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite 2% #51 of 134+1 versions & slices 2.1 Older, easier version of Terminal-Bench; Still widely reported AA 65.2% #43 of 156
DeepSWE1.1 Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition 9.1% #36 of 44+1 versions & slices 1 Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition Datacurve 20.4% #10 of 20
FrontierCode1.1 Agents implement changes in real repositories, graded against maintainer-defined requirements AA 23% #31 of 38
WebDev Arena Human pairwise preferences in the stated webdev vote pool Arena 1,482 #48 of 97+2 versions & slices October 23, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,482 #49 of 98October 19, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,482 #48 of 97
Terminal-Bench Hard Agents solve difficult software, system-administration and data-processing tasks in a terminal AA 42.4% #28 of 242
Agents 5
AutomationBench-AA Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation AA 21.3% #60 of 135
AA-Briefcase1.1 · Elo Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 1,089 #51 of 142
EnterpriseOps-Gym-AA Agents complete stateful enterprise workflows in eight domains; final database state determines success AA 32.1% #32 of 43
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 1,250 #48 of 177
τ³-BankingAA Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 15.3% #68 of 134
Multimodal 2
MMMU-Pro (vision) College-level questions that require understanding images: charts, diagrams, photos AA 78.6% #30 of 128
Arena Vision Human pairwise preferences in the stated vision vote pool Arena 1,251 #44 of 88
Writing 18
Arena Textcreative writing · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,403 #60 of 197+16 versions & slices creative writing Human pairwise preferences in the stated text vote pool Arena 1,405 #62 of 195overall Human pairwise preferences in the stated text vote pool Arena 1,441 #63 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,465 #69 of 198English Human pairwise preferences in the stated text vote pool Arena 1,442 #61 of 205French Human pairwise preferences in the stated text vote pool Arena 1,443 #70 of 164German Human pairwise preferences in the stated text vote pool Arena 1,430 #53 of 167instruction following Human pairwise preferences in the stated text vote pool Arena 1,432 #52 of 205Japanese Human pairwise preferences in the stated text vote pool Arena 1,379 #65 of 152Korean Human pairwise preferences in the stated text vote pool Arena 1,372 #68 of 156non-English Human pairwise preferences in the stated text vote pool Arena 1,419 #63 of 205Polish Human pairwise preferences in the stated text vote pool Arena 1,425 #60 of 142Russian Human pairwise preferences in the stated text vote pool Arena 1,427 #62 of 203Spanish Human pairwise preferences in the stated text vote pool Arena 1,432 #66 of 168overall · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,433 #61 of 188creative writing · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,404 #53 of 187overall · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,432 #65 of 198
Lech Mazur Short-Story Judges compare stories written to ten required elements; the comparison score can be negative Lech -0.1 #26 of 45
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 4.4% #84 of 103
Vals LegalBench Legal reasoning and legal-language understanding on LegalBench Vals 85.4% #17 of 89
Vals MedCode Medical coding accuracy on patient encounters Vals 46.3% #30 of 78
Vals MedScribe Accuracy of clinical notes produced from consultation transcripts Vals 87.3% #17 of 80
Vals Legal Research Bench Grounded legal research; a task passes only if all criteria pass Vals 29.8% #43 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 83% #17 of 317
GDP.pdfAA all-pass Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes 9.8% #64 of 135
Arena Document Human pairwise preferences in the stated document vote pool Arena 1,435 #26 of 33
Legacy · saturated or no longer reported 10
GPQA Diamond Graduate-level biology, chemistry and physics questions that can’t be solved by a quick web search 92.9% #13 of 315+1 versions & slices Epoch Models answer graduate-level science multiple-choice questions Epoch 81.3% #68 of 134
OTIS Mock AIME 20242025 Models solve the stated OTIS mock AIME mathematics problems Epoch 26.7% #102 of 131
SWE-rebenchall · March 1, 2026 to July 1, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 46.4% #2 of 6+5 versions & slices Go · March 1, 2026 to July 1, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 48.6% #3 of 6Java · March 1, 2026 to July 1, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 43.8% #2 of 6Python · March 1, 2026 to July 1, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 43.7% #2 of 6Rust · March 1, 2026 to July 1, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 46.4% #2 of 6TypeScript · March 1, 2026 to July 1, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 61.7% #2 of 6
τ²-benchAA Tests service agents that use tools while coordinating with a simulated user AA 88.9% #47 of 245
No benchmark matches
CheaperClaude Haiku 5.5max×3.8-62%$0.21Remove sampling parameters+5
BetterGPT-5.6 Lunalow×2.1-14%$0.01
OpenDeepSeek V4.1 Flash 1017max×6.1same$0.26Memory (4-bit): 256 GB → 458 GB+5
CheaperQwen 3.5 Flashunstated×2.5-78%$0.065 / $0.26
GLM 5.3maxOpen×4.1+310%$2.01Memory (4-bit): 256 GB → 452 GB+5
Nemotron 3 UltraunstatedOpen×2.5+100%$0.60 / $2.40Memory (4-bit): 256 GB → 336 GB+1