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 13.3 #162 of 339
Knowledge 5
AA OmniscienceAccuracy Measures correct answers on the AA Omniscience factual knowledge test AA 29.5% #82 of 323+1 versions & slices Hallucination Rate Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test AA 90.6% #255 of 323
AA Omniscience Index Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty AA -34.5 #160 of 323
IFBench (instructions) Checks whether the model follows precise formatting and content instructions AA 43.2% #140 of 248
SimpleQA Verifiedv1 Accuracy on short factual questions with verified answers, answered without web access 34.9% #48 of 84
Reasoning 4
ARC-AGI-1public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 77.1% #40 of 64+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 72.8% #42 of 67
ARC-AGI-2public evaluation Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 18.3% #40 of 66+1 versions & slices semi-private Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit ARC 17.6% #45 of 69
Math 16
MathArenaAIME 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 94.2% #7 of 30+15 versions & slices Apex Answer accuracy on the named mathematics competition under the recorded effort MathArena 1% #25 of 34Apex Shortlist Answer accuracy on the named mathematics competition under the recorded effort MathArena 55.9% #21 of 30BRUMO 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 93.3% #7 of 24CMIMC 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 94.4% #3 of 22Project Euler Answer accuracy on the named mathematics competition under the recorded effort MathArena 61.7% #7 of 11HMMT February 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 93.3% #5 of 30HMMT November 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 91.7% #5 of 14Kangaroo 2025 1-2 Answer accuracy on the named mathematics competition under the recorded effort MathArena 65.6% #12 of 17Kangaroo 2025 11-12 Answer accuracy on the named mathematics competition under the recorded effort MathArena 91.7% #10 of 17Kangaroo 2025 3-4 Answer accuracy on the named mathematics competition under the recorded effort MathArena 65.6% #10 of 17Kangaroo 2025 5-6 Answer accuracy on the named mathematics competition under the recorded effort MathArena 61.7% #16 of 17Kangaroo 2025 7-8 Answer accuracy on the named mathematics competition under the recorded effort MathArena 85.8% #13 of 16Kangaroo 2025 9-10 Answer accuracy on the named mathematics competition under the recorded effort MathArena 90.8% #15 of 16Putnam 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 48.3% #3 of 3SMT 2025 Answer accuracy on the named mathematics competition under the recorded effort MathArena 91% #4 of 24
Science 2
Humanity's Last Exam Very hard expert questions across many fields, written to stump today’s models AA 5.3% #268 of 339
CritPt (physics) Research-level physics problems that need many careful reasoning steps AA 0% #181 of 326
Coding 4
WebDev Arena Human pairwise preferences in the stated webdev vote pool Arena 1,396 #72 of 97+2 versions & slices October 23, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,396 #73 of 98October 19, 2026 Human pairwise preferences in the stated webdev vote pool Arena 1,396 #72 of 97
Terminal-Bench Hard Agents solve difficult software, system-administration and data-processing tasks in a terminal AA 22.7% #83 of 242
Agents 4
GDPval-AA (work tasks)2.1 Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work AA 16.3 #139 of 177
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 53.6% #14 of 16
MCP-Atlasfull 1000 Tool orchestration across MCP servers on the stated task split Scale 50.1% #14 of 16
τ³-BankingAA Agents retrieve banking knowledge and execute tools to resolve customer-support workflows AA 15.9% #66 of 134
Multimodal 3 Leads on 1
MMMU-Pro (vision) College-level questions that require understanding images: charts, diagrams, photos AA 62.4% #92 of 128
MMMUvalidation Image-grounded university-level questions across 30 disciplines; validation and test sets remain separate MMMU 85.4% #1 of 41
Arena Vision Human pairwise preferences in the stated vision vote pool Arena 1,250 #45 of 88
Writing 17
Arena Textcreative writing · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,427 #42 of 197+16 versions & slices creative writing Human pairwise preferences in the stated text vote pool Arena 1,429 #46 of 195overall Human pairwise preferences in the stated text vote pool Arena 1,456 #43 of 196Chinese Human pairwise preferences in the stated text vote pool Arena 1,495 #41 of 198English Human pairwise preferences in the stated text vote pool Arena 1,451 #48 of 205French Human pairwise preferences in the stated text vote pool Arena 1,450 #61 of 164German Human pairwise preferences in the stated text vote pool Arena 1,438 #49 of 167instruction following Human pairwise preferences in the stated text vote pool Arena 1,443 #41 of 205Japanese Human pairwise preferences in the stated text vote pool Arena 1,454 #19 of 152Korean Human pairwise preferences in the stated text vote pool Arena 1,400 #44 of 156non-English Human pairwise preferences in the stated text vote pool Arena 1,431 #45 of 205Polish Human pairwise preferences in the stated text vote pool Arena 1,440 #37 of 142Russian Human pairwise preferences in the stated text vote pool Arena 1,435 #50 of 203Spanish Human pairwise preferences in the stated text vote pool Arena 1,433 #65 of 168overall · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,443 #45 of 188creative writing · October 15, 2026 Human pairwise preferences in the stated text vote pool Arena 1,427 #42 of 187overall · October 22, 2026 Human pairwise preferences in the stated text vote pool Arena 1,443 #47 of 198
Health & law 3
Vals LegalBench Legal reasoning and legal-language understanding on LegalBench Vals 85.7% #16 of 89
Vals MedCode Medical coding accuracy on patient encounters Vals 52.7% #11 of 78
Vals MedScribe Accuracy of clinical notes produced from consultation transcripts Vals 88.1% #13 of 80
Documents 2
AA-LCR (long context)1.1 Questions that need reading and connecting facts across very long documents (around 100K tokens) AA 45% #188 of 317
Arena Document Human pairwise preferences in the stated document vote pool Arena 1,403 #33 of 33
Legacy · saturated or no longer reported 5
AIME2025 · AA archived AA evaluates the 30 integer-answer problems from the 2025 AIME mathematics competition AA 38% #98 of 140
LiveCodeBenchAA archived AA evaluates solutions to newly collected competitive-programming problems AA 49.4% #90 of 153
GPQA Diamond Graduate-level biology, chemistry and physics questions that can’t be solved by a quick web search 64.3% #214 of 315
τ²-benchAA Tests service agents that use tools while coordinating with a simulated user AA 46.5% #117 of 245
Terminal-Bench2.1 Older, easier version of Terminal-Bench; Still widely reported AA 52.4% #66 of 156
No benchmark matches
CheaperGemini 3 Flash Previewthinking×2.6-67%$0.50 / $3
OpenDeepSeek V3.2 Specialethinking×2.5——Context: 400K → 128K+1
CheaperGLM 4.7thinkingOpen×2.5-71%$0.60 / $2.20Context: 400K → 203K+1
CheaperKimi K2 ThinkingthinkingOpen×2.5-75%$0.47 / $2Context: 400K → 256K+1
Nova 2.0 Litehigh×2.5-75%$0.30 / $2.50No strict JSON
Nemotron 3 Nano 30B A3BthinkingOpen×2.4-97%$0.017Context: 400K → 131K+1
Qwen3 30B A3B 2507 (Reasoning)thinkingOpen+48%—$0.066Context: 400K → 262K+1