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 26.1 #75 of 337
Knowledge 6
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 23.7% #124 of 321+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 29.9% #34 of 321
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 0.9 #65 of 321
IFBench (instructions) Checks whether the model follows precise formatting and content instructions AA 76.3% #14 of 248
EQ-Bench3 Judge-rated responses to social and emotionally nuanced scenarios, scored through the source pairwise ranking EQ-Bench 1,567 #10 of 50
SimpleQA Verifiedv1 Accuracy on short factual questions with verified answers, answered without web access 34% #51 of 83
Reasoning 1
Chess Puzzles Models choose the best next move from a novel chess position Epoch 20% #54 of 68
Math 18
MathArenaAIME 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 95.8% #12 of 24+14 versions & slices HMMT February 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 89.4% #13 of 24Apex Answer accuracy on the named mathematics competition under the recorded effort MathArena 11.5% #17 of 34Apex Shortlist Answer accuracy on the named mathematics competition under the recorded effort MathArena 71.8% #12 of 30ArXivMath · April 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 36.3% #15 of 17ArXivMath · February 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 39.1% #13 of 20ArXivMath · January 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 65.2% #8 of 18ArXivMath · March 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 49.2% #11 of 14ArXivMath · May 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 52.5% #16 of 19BrokenArXiv · April 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 8.2% #14 of 15BrokenArXiv · February 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 9.3% #12 of 15BrokenArXiv · March 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 6.5% #11 of 13BrokenArXiv · May 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 10% #16 of 18ArXivLean · March 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 2.4% #5 of 5Project Euler Answer accuracy on the named mathematics competition under the recorded effort MathArena 67.2% #5 of 11
FrontierMath Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 33.4% #11 of 40
FrontierMath Tier4 Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 12.5% #10 of 33
FrontierMath Tiers 1–3 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 36.8% #52 of 69
Science 3
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 30.1% #80 of 337
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 4.6% #81 of 324
SciCode Writing code for real scientific computing problems (physics, biology, math) AA 44.8% #75 of 140
Coding 8
Terminal-Bench4.0 Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite 2% #49 of 132+1 versions & slices 2.1 Older, easier version of Terminal-Bench; Still widely reported AA 61.8% #50 of 156
DeepSWE1.1 Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition 18.9% #34 of 44+1 versions & slices 1 Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition Datacurve 17.5% #13 of 20
ITBench SRE IT incident tasks: diagnose and fix broken systems like a site-reliability engineer AA 40.3% #17 of 37
WebDev Arena Human pairwise preferences in the stated webdev vote pool Arena 1,508 #45 of 97+1 versions & slices full cohort Human pairwise preferences in the stated webdev vote pool Arena 1,508 #45 of 97
Terminal-Bench Hard Agents solve difficult software, system-administration and data-processing tasks in a terminal AA 43.2% #25 of 242
Agents 4
AutomationBench-AA Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation AA 20.3% #60 of 133
AA-Briefcase1.1 · Elo Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 965 #58 of 140
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 30.8 #62 of 175
τ³-BankingAA Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 13.6% #75 of 134
Writing 18
Arena Textcreative writing Human pairwise preferences in the stated text vote pool Arena 1,447 #29 of 195+14 versions & slices overall Human pairwise preferences in the stated text vote pool Arena 1,465 #36 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,514 #26 of 181English Human pairwise preferences in the stated text vote pool Arena 1,475 #23 of 188French Human pairwise preferences in the stated text vote pool Arena 1,472 #30 of 149German Human pairwise preferences in the stated text vote pool Arena 1,467 #24 of 154instruction following Human pairwise preferences in the stated text vote pool Arena 1,452 #35 of 188Japanese Human pairwise preferences in the stated text vote pool Arena 1,436 #25 of 138Korean Human pairwise preferences in the stated text vote pool Arena 1,422 #29 of 144non-English Human pairwise preferences in the stated text vote pool Arena 1,448 #31 of 188Polish Human pairwise preferences in the stated text vote pool Arena 1,447 #30 of 133Russian Human pairwise preferences in the stated text vote pool Arena 1,454 #32 of 186Spanish Human pairwise preferences in the stated text vote pool Arena 1,469 #21 of 155overall · full cohort Human pairwise preferences in the stated text vote pool Arena 1,462 #28 of 188creative writing · full cohort Human pairwise preferences in the stated text vote pool Arena 1,454 #23 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 73.5% #30 of 95
EQ-Bench Creative Writingv3 Creative Writing v3 · Elo: Judge-rated creative writing under the v3 prompt set and pairwise ranking EQ-Bench 1,592 #43 of 91
Lech Mazur Short-Story Judges compare stories written to ten required elements; the comparison score can be negative Lech -1 #30 of 44
Health & law 4
Vals LegalBench Legal reasoning and legal-language understanding on LegalBench Vals 84.4% #29 of 89
Vals MedCode Medical coding accuracy on patient encounters Vals 41.6% #44 of 78
Vals MedScribe Accuracy of clinical notes produced from consultation transcripts Vals 72.3% #66 of 80
Vals Legal Research Bench Grounded legal research; a task passes only if all criteria pass Vals 27.9% #44 of 69
Documents 2
GDP.pdfAA all-pass Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes 8.4% #66 of 133
AA-LCR (long context)1.1 Questions that need reading and connecting facts across very long documents (around 100K tokens) AA 73.7% #97 of 275
Legacy · saturated or no longer reported 17
GPQA Diamond Graduate-level biology, chemistry and physics questions that can’t be solved by a quick web search 86.8% #61 of 315+1 versions & slices Epoch Models answer graduate-level science multiple-choice questions Epoch 89.9% #34 of 133
OTIS Mock AIME 20242025 Models solve the stated OTIS mock AIME mathematics problems Epoch 93.3% #38 of 130
SWE-bench VerifiedEpoch Swe Bench Verified: Agents modify real repositories to resolve the stated Verified issues Epoch 74.2% #13 of 24
SWE-rebenchall · 2026-02-01 to 2026-03-01 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 62.7% #1 of 5+11 versions & slices all · 2026-03-01 to 2026-05-15 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 50.7% #3 of 5Go · 2026-02-01 to 2026-03-01 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5Go · 2026-03-01 to 2026-05-15 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5Java · 2026-02-01 to 2026-03-01 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5Java · 2026-03-01 to 2026-05-15 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5Python · 2026-02-01 to 2026-03-01 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 62.7% #1 of 5Python · 2026-03-01 to 2026-05-15 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 50.7% #3 of 5Rust · 2026-02-01 to 2026-03-01 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5Rust · 2026-03-01 to 2026-05-15 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5TypeScript · 2026-02-01 to 2026-03-01 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5TypeScript · 2026-03-01 to 2026-05-15 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5
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×2.2-79%$0.18
CheaperClaude Haiku 5.5max×2-91%$0.21Remove sampling parameters+5
CheaperMuse Spark 1.3xhigh×2-7%$1.37Only auto tool_choice is supported+5
Grok 4.3high+16%-27%$0.21
DeepSeek V4 Pro 0813maxOpen+75%-8%$0.67Memory (4-bit): 452 GB → 990 GB+5
Kimi K2.6unstatedOpen+55%-20%$0.95 / $4Memory (4-bit): 452 GB → 616 GB