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 41.8 #28 of 339
Knowledge 7
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 48.8% #28 of 323+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 39.3% #60 of 323
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 28.8 #15 of 323
IFBench (instructions) Checks whether the model follows precise formatting and content instructions AA 62.2% #80 of 248
EQ-Bench3 Judge-rated responses to social and emotionally nuanced scenarios, scored through the source pairwise ranking EQ-Bench 2,058 #2 of 53
SimpleQA Verifiedv1 Accuracy on short factual questions with verified answers, answered without web access 53% #21 of 84
Mystery Game Puzzles Models choose the best move in puzzle positions from an undisclosed game Epoch 36% #17 of 65
Reasoning 4
ARC-AGI-1semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 91.5% #22 of 67
ARC-AGI-2semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 71.7% #21 of 69
ARC-AGI-3semi-private Measures interactive game solving through relative human action efficiency ARC 1.5% #11 of 19
Chess Puzzles Models choose the best next move from a novel chess position Epoch 52% #13 of 69
Math 26 Leads on 2
AA Analyst Agent Data-analyst tasks: the model explores data with tools and reports findings AA 45% #15 of 43
MathArenaAIME 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 100% #1 of 24+20 versions & slices HMMT February 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 95.5% #6 of 24Apex Answer accuracy on the named mathematics competition under the recorded effort MathArena 81.3% #1 of 34Apex Shortlist Answer accuracy on the named mathematics competition under the recorded effort MathArena 90.4% #4 of 30ArXivMath · April 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 62.5% #8 of 17ArXivMath · February 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 60.2% #4 of 20ArXivMath · June 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 70% #13 of 24ArXivMath · March 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 75% #2 of 14ArXivMath · May 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 65% #9 of 19BrokenArXiv · April 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 34.4% #6 of 15BrokenArXiv · February 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 35.5% #3 of 15BrokenArXiv · June 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 41.7% #15 of 24BrokenArXiv · March 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 35.7% #3 of 13BrokenArXiv · May 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 23% #11 of 18ArXivLean · June 2026 Answer accuracy on the named mathematics competition under the recorded effort MathArena 23.9% #4 of 9Kangaroo 2025 1-2 Answer accuracy on the named mathematics competition under the recorded effort MathArena 70.8% #10 of 17Kangaroo 2025 11-12 Answer accuracy on the named mathematics competition under the recorded effort MathArena 100% #1 of 17Kangaroo 2025 3-4 Answer accuracy on the named mathematics competition under the recorded effort MathArena 60.4% #14 of 17Kangaroo 2025 5-6 Answer accuracy on the named mathematics competition under the recorded effort MathArena 70% #10 of 17Kangaroo 2025 7-8 Answer accuracy on the named mathematics competition under the recorded effort MathArena 96.7% #1 of 16Kangaroo 2025 9-10 Answer accuracy on the named mathematics competition under the recorded effort MathArena 91.7% #13 of 16
FrontierMath Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 47.2% #3 of 40
FrontierMath Tier4 Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 31.3% #1 of 33
FrontierMath Tier 4 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 56.1% #17 of 55
FrontierMath Tiers 1–3 (v2) Models solve expert-written advanced mathematics problems on the named tier and edition Epoch 80% #17 of 70
Science 3
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 48.7% #14 of 339
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 20.9% #27 of 326
SciCode Writing code for real scientific computing problems (physics, biology, math) AA 54.4% #33 of 142
Coding 8
Terminal-Bench4.0 Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite 21.7% #30 of 134+1 versions & slices 2.1 Older, easier version of Terminal-Bench; Still widely reported AA 84.6% #12 of 156
DeepSWE1.1 Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition 48.7% #27 of 44+1 versions & slices 1 Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition Datacurve 47.3% #4 of 20
WebDev Arena Human pairwise preferences in the stated webdev vote pool Arena 1,556 #33 of 97+2 versions & slices October 23, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,556 #34 of 98October 19, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,556 #33 of 97
Terminal-Bench Hard Agents solve difficult software, system-administration and data-processing tasks in a terminal AA 58.3% #5 of 242
Agents 9
AutomationBench-AA Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation AA 45.6% #43 of 135
AA-Briefcase1.1 · Elo Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 1,319 #38 of 142
EnterpriseOps-Gym-AA Agents complete stateful enterprise workflows in eight domains; final database state determines success AA 44% #14 of 43
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 1,454 #36 of 177
Toolathlon-Verified Toolathlon-Verified as reported in this launch evaluation; the post is the authority for the recorded task and grading scope 76.2% #3 of 6
τ³-benchbanking · 1.0.1 τ-bench official · 3-banking-1.0.1: Customer-service agents converse with simulated users, retrieve domain knowledge and change state through tools Sierra 39.7% #7 of 21
EBR Bench Agents learn to play Earthborne Rangers across repeated playthroughs Epoch 28.6% #14 of 24
Furniture Assembly Models inspect furniture assembly photos, identify incorrect builds and locate the mistaken step Epoch 42.5% #15 of 31
τ³-BankingAA Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 34.2% #30 of 134
Multimodal 2
Chartography Expert chart reasoning over professional scientific, medical, financial and engineering charts Surge 15.9% #28 of 37
Arena Vision Human pairwise preferences in the stated vision vote pool Arena 1,294 #15 of 88
Writing 20
Arena Textcreative writing · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,454 #21 of 197+16 versions & slices creative writing Human pairwise preferences in the stated text vote pool Arena 1,469 #12 of 195overall Human pairwise preferences in the stated text vote pool Arena 1,481 #19 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,506 #34 of 198English Human pairwise preferences in the stated text vote pool Arena 1,469 #29 of 205French Human pairwise preferences in the stated text vote pool Arena 1,481 #25 of 164German Human pairwise preferences in the stated text vote pool Arena 1,473 #18 of 167instruction following Human pairwise preferences in the stated text vote pool Arena 1,476 #19 of 205Japanese Human pairwise preferences in the stated text vote pool Arena 1,442 #25 of 152Korean Human pairwise preferences in the stated text vote pool Arena 1,430 #23 of 156non-English Human pairwise preferences in the stated text vote pool Arena 1,450 #29 of 205Polish Human pairwise preferences in the stated text vote pool Arena 1,459 #23 of 142Russian Human pairwise preferences in the stated text vote pool Arena 1,474 #20 of 203Spanish Human pairwise preferences in the stated text vote pool Arena 1,465 #25 of 168overall · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,461 #29 of 188creative writing · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,454 #21 of 187overall · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,461 #29 of 198
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 80.8% #15 of 96
EQ-Bench Creative Writingv3 Creative Writing v3 · Elo: Judge-rated creative writing under the v3 prompt set and pairwise ranking EQ-Bench 1,840 #22 of 94
Lech Mazur Short-Story Judges compare stories written to ten required elements; the comparison score can be negative Lech 0.7 #16 of 45
Health & law 5
MLCR-AA Models synthesize long medical records; the headline credits concise answers judged both accurate and complete AA 19.4% #26 of 103
Vals LegalBench Legal reasoning and legal-language understanding on LegalBench Vals 83.6% #44 of 89
Vals MedCode Medical coding accuracy on patient encounters Vals 53.2% #8 of 78
Vals MedScribe Accuracy of clinical notes produced from consultation transcripts Vals 85.8% #24 of 80
Vals Legal Research Bench Grounded legal research; a task passes only if all criteria pass Vals 43.8% #19 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 77.7% #75 of 317
GDP.pdfAA all-pass Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes 22.8% #16 of 135
Arena Document Human pairwise preferences in the stated document vote pool Arena 1,475 #9 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% #21 of 315+1 versions & slices Epoch Models answer graduate-level science multiple-choice questions Epoch 91% #24 of 134
OTIS Mock AIME 20242025 Models solve the stated OTIS mock AIME mathematics problems Epoch 98.3% #21 of 131
SWE-rebenchall · March 1, 2026 to May 15, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 56.5% #1 of 5+5 versions & slices Go · March 1, 2026 to May 15, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5Java · March 1, 2026 to May 15, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5Python · March 1, 2026 to May 15, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 56.5% #1 of 5Rust · March 1, 2026 to May 15, 2026 Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 5TypeScript · March 1, 2026 to May 15, 2026 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 94.4% #22 of 245
No benchmark matches
CheaperGemini 3.8 Flashhigh+8%-85%*$3.49Sampling parameters are ignored+1
BetterGPT-6.1 Sollow+2%-60%$0.13Remove unsupported parameters+5
OpenDeepSeek V4.1 Flash 1017high+3%-95%$0.30 / $1.20Pass back reasoning_content · Chat Completions+4
Kimi K3maxOpen+3%-40%$2.00Sampling parameters are fixed+3
GLM 5.3 FlashmaxOpensame-98%$0.28Defaults to max · Self-hosted+4
Muse Spark OpenxhighOpen-3%-80%$0.84Reasoning off leaves output unchanged · Self-hosted / llama.cpp+1