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 30.1 #57 of 337
Knowledge 7
AA OmniscienceAccuracy AA Omniscience · Accuracy: Measures correct answers on the AA Omniscience factual knowledge test AA 40.9% #45 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 48.4% #81 of 321
AA Omniscience Index AA Omniscience Index: Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA 12.2 #36 of 321
IFBench (instructions) IFBench (instructions): Checks whether the model follows precise formatting and content instructions AA 56.6% #99 of 248
EQ-Bench3 EQ-Bench 3: Judge-rated responses to social and emotionally nuanced scenarios, scored through the source pairwise ranking EQ-Bench 1,714 #5 of 50
SimpleQA Verifiedv1 SimpleQA Verified v1: Accuracy on short factual questions with verified answers, answered without web access 35.5% #46 of 83
Mystery Game Puzzles Mystery Game Puzzles: Models choose the best move in puzzle positions from an undisclosed game Epoch 16% #49 of 64
Reasoning 5
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 95.8% #21 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 86.5% #32 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 65.7% #20 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 60.4% #26 of 66
Chess Puzzles Chess Puzzles: Models choose the best next move from a novel chess position Epoch 16% #56 of 68
Math 1
AA Analyst Agent AA Analyst Agent: Data-analyst tasks: the model explores data with tools and reports findings AA 20% #30 of 43
Science 3
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models AA 33.6% #70 of 337
CritPt (physics) CritPt (physics): Research-level physics problems that need many careful reasoning steps AA 3.1% #92 of 324
SciCode SciCode: Writing code for real scientific computing problems (physics, biology, math) AA 50.1% #50 of 140
Coding 8
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 3% #46 of 132+1 versions & slices 2.1 Terminal-Bench 2.1 · source setting 2: Older, easier version of Terminal-Bench; Still widely reported AA 71.2% #34 of 156
DeepSWE1.1 DeepSWE 1.1: Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition 29.9% #33 of 44+1 versions & slices 1 DeepSWE 1: Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition Datacurve 31.8% #6 of 20
ITBench SRE ITBench SRE: IT incident tasks: diagnose and fix broken systems like a site-reliability engineer AA 39.8% #19 of 37
WebDev Arena WebDev Arena: Human pairwise preferences in the stated webdev vote pool Arena 1,523 #42 of 97+1 versions & slices full cohort WebDev Arena · snapshot 2026-10-19 · native board settings: Human pairwise preferences in the stated webdev vote pool Arena 1,523 #42 of 97
Terminal-Bench Hard Terminal-Bench Hard (version unstated): Agents solve difficult software, system-administration and data-processing tasks in a terminal AA 53% #9 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 20.1% #61 of 133
AA-Briefcase1.1 · Elo AA-Briefcase 1.1 · Elo: Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating AA 1,063 #51 of 140
APEX-Agents APEX-Agents: Agents complete long-horizon professional-services tasks across realistic applications AA 28% #14 of 29
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 36.6 #50 of 175
MCP-Atlas Public MCPAtlas Public: MCPAtlas Public as reported in this launch evaluation; the post is the authority for the recorded task and grading scope 72.8% #8 of 16
MCP-Atlasfull 1000 MCP-Atlas · full 1000: Tool orchestration across MCP servers on the stated task split Scale 69.5% #8 of 16
τ³-BankingAA τ³-Banking · AA: Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 34.4% #28 of 134
Multimodal 2
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos AA 73.3% #62 of 127
Arena Vision Arena Vision · overall: Human pairwise preferences in the stated vision vote pool Arena 1,283 #22 of 88
Writing 17
Arena Textcreative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool Arena 1,449 #27 of 195+14 versions & slices overall Arena Text · overall: Human pairwise preferences in the stated text vote pool Arena 1,472 #31 of 196Chinese Arena Text · chinese: Human pairwise preferences in the stated text vote pool Arena 1,492 #49 of 181English Arena Text · english: Human pairwise preferences in the stated text vote pool Arena 1,472 #26 of 188French Arena Text · french: Human pairwise preferences in the stated text vote pool Arena 1,465 #39 of 149German Arena Text · german: Human pairwise preferences in the stated text vote pool Arena 1,428 #50 of 154instruction following Arena Text · instruction following: Human pairwise preferences in the stated text vote pool Arena 1,476 #20 of 188Japanese Arena Text · japanese: Human pairwise preferences in the stated text vote pool Arena 1,420 #33 of 138Korean Arena Text · korean: Human pairwise preferences in the stated text vote pool Arena 1,411 #35 of 144non-English Arena Text · non english: Human pairwise preferences in the stated text vote pool Arena 1,440 #34 of 188Polish Arena Text · polish: Human pairwise preferences in the stated text vote pool Arena 1,442 #32 of 133Russian Arena Text · russian: Human pairwise preferences in the stated text vote pool Arena 1,439 #43 of 186Spanish Arena Text · spanish: Human pairwise preferences in the stated text vote pool Arena 1,464 #26 of 155overall · full cohort Arena Text · overall · snapshot 2026-10-15 · native board settings: Human pairwise preferences in the stated text vote pool Arena 1,458 #30 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 1,435 #37 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 79.9% #15 of 95
EQ-Bench Creative Writingv3 Creative Writing v3 · Elo: Judge-rated creative writing under the v3 prompt set and pairwise ranking EQ-Bench 1,810 #23 of 91
Health & law 3
MLCR-AA MLCR-AA: Models synthesize long medical records; the headline credits concise answers judged both accurate and complete AA 24.4% #19 of 64
Vals LegalBench Vals LegalBench: Legal reasoning and legal-language understanding on LegalBench Vals 82.1% #54 of 89
Vals Legal Research Bench Vals Legal Research Bench: Grounded legal research; a task passes only if all criteria pass Vals 38.5% #28 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 80% #42 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 15.8% #36 of 133
Arena Document Arena Document · overall: Human pairwise preferences in the stated document vote pool Arena 1,482 #8 of 33
Legacy · saturated or no longer reported 11
GPQA Diamond GPQA Diamond: Graduate-level biology, chemistry and physics questions that can’t be solved by a quick web search 87.5% #52 of 315+1 versions & slices Epoch GPQA Diamond · source setting 2: Models answer graduate-level science multiple-choice questions Epoch 83.3% #60 of 133
OTIS Mock AIME 20242025 OTIS Mock AIME 2024 2025: Models solve the stated OTIS mock AIME mathematics problems Epoch 82.2% #68 of 130
SWE-bench VerifiedEpoch Swe Bench Verified: Agents modify real repositories to resolve the stated Verified issues Epoch 75.2% #11 of 24
SWE-rebenchall · 2026-02-01 to 2026-05-15 SWE-rebench · all · 2026-02-01 to 2026-05-15: Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 54.5% #3 of 4+5 versions & slices Go · 2026-02-01 to 2026-05-15 SWE-rebench · go · 2026-02-01 to 2026-05-15: Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 4Java · 2026-02-01 to 2026-05-15 SWE-rebench · java · 2026-02-01 to 2026-05-15: Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 4Python · 2026-02-01 to 2026-05-15 SWE-rebench · python · 2026-02-01 to 2026-05-15: Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 54.5% #3 of 4Rust · 2026-02-01 to 2026-05-15 SWE-rebench · rust · 2026-02-01 to 2026-05-15: Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 4TypeScript · 2026-02-01 to 2026-05-15 SWE-rebench · typescript · 2026-02-01 to 2026-05-15: Repository issue resolution over the exact fresh pull-request window, under the source agent version Nebius 0% #1 of 4
τ²-benchAA τ²-Bench: Tests service agents that use tools while coordinating with a simulated user AA 79.5% #82 of 245
No benchmark matches
BetterGPT-6 Lunahigh+10%-99%$0.029Function calling only with none · Chat Completions+3
OpenMiMo v2.6 Prothinking+54%-95%$0.13Pass back reasoning_content+3
CheaperGemini 4 Argonhigh+75%-20%*$1.99Synchronous output stops at 65,536+2
DeepSeek V4.1 Flash 1017maxOpen+50%-89%$0.26Pass back reasoning_content · Chat Completions+4
GLM 5.3maxOpen+49%-19%$2.01No image input+4
Step 5 PreviewthinkingOpen+46%-59%$1.03web_search not available · Chat Completions+2
Kimi K3maxOpen+45%-20%$2.00Sampling parameters are fixed+3