LLM benchmark leaderboards

Evals

As of August 20, 2026, this index ranks 696 public LLM evals next to live API token prices. Every score on a model or comparison page links here. Official boards own the methodology; these pages add $/M.

One page per public eval with at least one published score. Sorted by US search demand, then by how many models have a score. Prices stay on the ranking page.
Benchmark	US searches / mo	Models scored	What it measures
HLE	6,600	104	Humanity's Last Exam: 2,500 expert questions designed so search engines fail.
SWE Bench	6,600	8	the SWE Bench score
Humanity's Last Exam (No Tools)	6,600	5	the Humanity's Last Exam (No Tools) score
Humanity's Last Exam (With Tools)	6,600	5	the Humanity's Last Exam (With Tools) score
Terminal-Bench 2.0	5,400	68	the Terminal-Bench 2.0 score
Terminal-Bench	5,400	67	Terminal-Bench: can the model finish jobs in a real shell.
Terminal-Bench 2.1	5,400	52	Terminal-Bench: can the model finish jobs in a real shell.
Terminal-Bench 2	5,400	45	the Terminal-Bench 2 score
Terminal-Bench 3.0	5,400	3	the Terminal-Bench 3.0 score
ARC-AGI-1 Verified	3,600	72	the ARC-AGI-1 Verified score
Chatbot Arena	3,600	16	Chatbot Arena human votes: people pick the better reply in a blind test. That is a writing and taste rank, not a trivia quiz.
ARC-AGI-3	3,600	10	the ARC-AGI-3 score
Arc Agi	3,600	8	the Arc Agi score
DeepSWE	2,900	34	DeepSWE: repository-level coding agents on real GitHub work.
SWE-bench Verified	2,400	137	the SWE-bench Verified score
SWE-bench Pro	2,400	52	the SWE-bench Pro score
SWE-bench Pro (Public)	2,400	10	the SWE-bench Pro (Public) score
MMLU	1,900	87	MMLU, a multiple-choice knowledge exam. It does not measure prose.
ARC-AGI 2	1,900	69	ARC-AGI: novel visual puzzles, not memorized exams.
LiveBench	1,900	55	the LiveBench score
OSWorld	1,900	38	OSWorld: can the model drive a real desktop (click, type, finish the task).
OSWorld Verified	1,900	36	OSWorld-Verified: cleaned desktop computer-use tasks with an execution check.
GDPval	1,900	16	GDPval: professional knowledge-work, closer to office output than trivia.
OSWorld 2.0	1,900	12	the OSWorld 2.0 score
Screenspot	1,900	11	the Screenspot score
ARC-AGI-2 Verified	1,900	2	the ARC-AGI-2 Verified score
ARC-AGI 2	1,900	1	the ARC-AGI 2 score
SimpleBench	1,600	59	the SimpleBench score
GSM8K	1,600	36	the GSM8K score
Treebench	1,600	1	the Treebench score
LiveCodeBench	1,300	128	LiveCodeBench: contest programming on problems released after training.
tau-bench Retail	1,300	44	tau-bench: multi-turn tool use on customer-support workflows.
LiveCodeBench v6	1,300	43	the LiveCodeBench v6 score
tau2-bench (Telecom)	1,300	31	the tau2-bench (Telecom) score
tau2-bench Telecom	1,300	30	the tau2-bench Telecom score
Tau Bench	1,300	6	the Tau Bench score
GPQA	1,000	189	GPQA: graduate-level science questions that resist simple search.
BrowseComp	1,000	59	the BrowseComp score
Tvbench	1,000	2	the Tvbench score
Deck Bench	1,000	1	the Deck Bench score
AIME 2025	880	170	AIME 2025: American invitational math contest problems.
GPQA Diamond	880	126	GPQA Diamond: graduate-level science questions written so Google cannot solve them.
Aime 2026	880	14	the Aime 2026 score
HealthBench	880	9	HealthBench: medical conversation quality, not multiple-choice science.
MMLU-Pro	720	165	MMLU-Pro, a hard multiple-choice knowledge quiz. It is not a writing test.
HumanEval	720	51	the HumanEval score
MMMU	590	98	MMMU: college-level questions on diagrams, charts, and images.
FrontierMath	590	53	FrontierMath: research-level math that stays hard after models ace AIME.
CursorBench	590	14	the CursorBench score
Big Bench	590	2	the Big Bench score
Mle Bench	590	2	the Mle Bench score
SimpleQA	480	86	SimpleQA, which checks whether the model makes up facts.
IFEval	480	49	IFEval, which checks whether the model followed the prompt (format, length, constraints). That is instruction-following, not 'sounds good.'
MATH 500	480	42	MATH-500: textbook and contest math problems.
SkillsBench	480	30	SkillsBench: can the agent actually use bundled skills and tools on multi-step tasks, not just chat about them.
HellaSwag	480	23	HellaSwag, a sentence-completion quiz, not a writing sample.
Mt Bench	480	8	the Mt Bench score
Omnidocbench	480	2	the Omnidocbench score
AIME 2024	390	41	the AIME 2024 score
Vending Bench 2	390	4	the Vending Bench 2 score
Pinchbench	390	3	the Pinchbench score
Eq Bench	390	2	the Eq Bench score
MMMU-Pro	320	54	MMMU-Pro: a harder multimodal exam than MMMU.
IFBench	320	25	IFBench, a harder instruction-following set than IFEval.
ScreenSpot Pro	320	21	the ScreenSpot Pro score
Mrcr	320	7	the Mrcr score
Paperbench	320	3	the Paperbench score
Cybench	320	2	the Cybench score
DeepResearch Bench	260	20	the DeepResearch Bench score
Triviaqa	260	7	the Triviaqa score
Gdpval Aa	260	6	the Gdpval Aa score
Mmbench	260	2	the Mmbench score
TriviaQA	260	2	the TriviaQA score
SciCode	210	90	the SciCode score
ToolAthlon	210	37	the ToolAthlon score
MathVista	210	30	the MathVista score
Ocrbench	210	16	the Ocrbench score
Longbench V2	210	12	the Longbench V2 score
Program Bench	210	5	the Program Bench score
Livecodebench Pro	210	3	the Livecodebench Pro score
Cl Bench	210	2	the Cl Bench score
Bixbench	210	1	the Bixbench score
MMMLU	170	38	the MMMLU score
SuperGPQA	170	28	the SuperGPQA score
Mvbench	170	13	the Mvbench score
Claw Eval	170	11	the Claw Eval score
Exploitbench	170	5	the Exploitbench score
Posttrainbench	170	5	the Posttrainbench score
Swe Marathon	170	4	the Swe Marathon score
SWE-bench Multilingual	140	38	the SWE-bench Multilingual score
MRCR v2	140	22	MRCR: can the model still find planted facts inside a long prompt.
LVBench	140	21	the LVBench score
WinoGrande	140	16	the WinoGrande score
Longvideobench	140	2	the Longvideobench score
Evalplus	140	1	the Evalplus score
FrontierMath Tier 4	110	40	the FrontierMath Tier 4 score
Fiction.liveBench	110	35	the Fiction.liveBench score
Big Bench Hard	110	13	the Big Bench Hard score
BBH	110	12	the BBH score
HealthBench Professional	110	9	the HealthBench Professional score
Finance Agent	110	8	the Finance Agent score
Zerobench	110	7	the Zerobench score
Multi Swe Bench	110	6	the Multi Swe Bench score
Biomysterybench	110	3	the Biomysterybench score
Ocrbench V2	110	3	the Ocrbench V2 score
Swe Lancer	110	3	the Swe Lancer score
Deep Swe	110	1	the Deep Swe score
Simpleqa Verified	110	1	the Simpleqa Verified score
Swe Atlas	110	1	the Swe Atlas score
MMLU-Redux	90	37	the MMLU-Redux score
Legal Agent Benchmark	90	14	the Legal Agent Benchmark score
Writingbench	90	14	the Writingbench score
Mmlongbench Doc	90	5	the Mmlongbench Doc score
Tau3 Bench	90	5	the Tau3 Bench score
Swe Bench Multimodal	90	4	the Swe Bench Multimodal score
Wildclawbench	90	4	the Wildclawbench score
Acebench	90	2	the Acebench score
Terminal Bench Hard	90	2	the Terminal Bench Hard score
Web Bench	90	1	the Web Bench score
ProofBench	70	52	the ProofBench score
CyberBench	70	35	the CyberBench score
VideoMMMU	70	25	the VideoMMMU score
T2 Bench	70	15	the T2 Bench score
AutomationBench	70	13	the AutomationBench score
C Eval	70	11	the C Eval score
Lifescibench	70	3	the Lifescibench score
Cfeval	70	2	the Cfeval score
Genebench	70	2	the Genebench score
Graphwalks	70	2	the Graphwalks score
Automation Bench	70	1	the Automation Bench score
Big Bench Audio	70	1	the Big Bench Audio score
Swt Bench	70	1	the Swt Bench score
Webdev Arena	70	1	the Webdev Arena score
Worldbench	70	1	the Worldbench score
APEX-Agents	50	50	the APEX-Agents score
Browsecomp Zh	50	12	the Browsecomp Zh score
HealthBench Hard	50	8	the HealthBench Hard score
Agieval	50	4	the Agieval score
Job Bench	50	4	the Job Bench score
Blueprint Bench 2	50	3	the Blueprint Bench 2 score
Genebench Pro	50	3	the Genebench Pro score
Global Mmlu	50	2	the Global Mmlu score
Motionbench	50	2	the Motionbench score
Labbench2	50	1	the Labbench2 score
Osworld G	50	1	the Osworld G score
Ovobench	50	1	the Ovobench score
Phybench	50	1	the Phybench score
Yc Bench	50	1	the Yc Bench score
Mmlu Prox	40	25	the Mmlu Prox score
Muirbench	40	9	the Muirbench score
Ojbench	40	9	the Ojbench score
Complexfuncbench	40	6	the Complexfuncbench score
Countbench	40	6	the Countbench score
Big Bench Extra Hard	40	5	the Big Bench Extra Hard score
Benchcad	40	4	the Benchcad score
Cmmlu	40	4	the Cmmlu score
Zclawbench	40	4	the Zclawbench score
Gdpval Aa V2	40	2	the Gdpval Aa V2 score
Bankertoolbench	40	1	the Bankertoolbench score
Posttrain Bench	40	1	the Posttrain Bench score
Vibe Code Bench	30	59	the Vibe Code Bench score
MCP Atlas	30	37	the MCP Atlas score
Imo Answerbench	30	20	the Imo Answerbench score
AA-LCR	30	15	long-context retrieval: can the model use a huge window, not just advertise one.
Vita Bench	30	9	the Vita Bench score
Tir Bench	30	4	the Tir Bench score
Sec Bench Pro	30	3	the Sec Bench Pro score
Workspace Bench	30	2	the Workspace Bench score
Cmt Benchmark	30	1	the Cmt Benchmark score
Frontier-Bench v0.1	30	1	the Frontier-Bench v0.1 score
Frontier Swe	30	1	the Frontier Swe score
Gdpval Aa Elo	30	1	the Gdpval Aa Elo score
Harvey Lab	30	1	the Harvey Lab score
Livesqlbench	30	1	the Livesqlbench score
Longcodebench	30	1	the Longcodebench score
Mle Bench Lite	30	1	the Mle Bench Lite score
Profbench	30	1	the Profbench score
Qwenclawbench	30	1	the Qwenclawbench score
Swe Perf	30	1	the Swe Perf score
DeepSWE 1.1	20	27	DeepSWE 1.1: repository-level coding agents on real GitHub work.
GDP-PDF	20	22	the GDP-PDF score
Arc C	20	21	the Arc C score
Tau Bench Airline	20	21	the Tau Bench Airline score
Arena-Hard	20	20	the Arena-Hard score
Bfcl V3	20	17	the Bfcl V3 score
Video Mme	20	15	the Video Mme score
Hallusion Bench	20	14	the Hallusion Bench score
Global Mmlu Lite	20	10	the Global Mmlu Lite score
Vibe Eval	20	8	the Vibe Eval score
Video-MME	20	7	the Video-MME score
Androidworld	20	5	the Androidworld score
Graphwalks BFS (256K-1M)	20	5	the Graphwalks BFS (256K-1M) score
Livecodebench V5	20	5	the Livecodebench V5 score
FrontierMath (Tiers 1-3)	20	4	the FrontierMath (Tiers 1-3) score
Alignbench	20	3	the Alignbench score
Big Finance Bench	20	3	the Big Finance Bench score
Lmarena Text	20	2	the Lmarena Text score
Multilingual Mmlu	20	2	the Multilingual Mmlu score
Visualwebbench	20	2	the Visualwebbench score
Browsecomp Vl	20	1	the Browsecomp Vl score
Deepswe V1 1	20	1	the Deepswe V1 1 score
Humaneval Plus	20	1	the Humaneval Plus score
Kernelbench Hard	20	1	the Kernelbench Hard score
Mmbench Video	20	1	the Mmbench Video score
Octocodingbench	20	1	the Octocodingbench score
Presentbench	20	1	the Presentbench score
Researchclawbench	20	1	the Researchclawbench score
Svg Bench	20	1	the Svg Bench score
Swe Fficiency	20	1	the Swe Fficiency score
HMMT 2025	10	28	the HMMT 2025 score
tau2-bench Retail	10	24	the tau2-bench Retail score
tau2-bench Airline	10	21	the tau2-bench Airline score
Mathvista Mini	10	19	the Mathvista Mini score
Realworldqa	10	19	the Realworldqa score
Multi If	10	18	the Multi If score
Cc Ocr	10	16	the Cc Ocr score
Mm Mt Bench	10	15	the Mm Mt Bench score
Graphwalks BFS (0K-128K)	10	14	the Graphwalks BFS (0K-128K) score
Arena Hard V2	10	13	the Arena Hard V2 score
Livebench 20241125	10	13	the Livebench 20241125 score
Mmbench V1 1	10	13	the Mmbench V1 1 score
Creative Writing V3	10	12	the Creative Writing V3 score
Facts Grounding	10	12	the Facts Grounding score
Mmmu Val	10	10	the Mmmu Val score
Agents' Last Exam	10	9	the Agents' Last Exam score
Bfcl V4	10	8	the Bfcl V4 score
Deep Planning	10	8	the Deep Planning score
Multipl E	10	8	the Multipl E score
Nova 63	10	8	the Nova 63 score
The Agent Company	10	8	the The Agent Company score
Wild Bench	10	8	the Wild Bench score
Embspatialbench	10	7	the Embspatialbench score
Matharena Apex	10	7	the Matharena Apex score
OfficeQA Pro	10	7	the OfficeQA Pro score
Refspatialbench	10	6	the Refspatialbench score
Seal 0	10	6	the Seal 0 score
Coworkbench	10	4	the Coworkbench score
HealthBench Consensus	10	4	the HealthBench Consensus score
Mrcr 1m	10	4	the Mrcr 1m score
Aa Briefcase	10	3	the Aa Briefcase score
Api Bank	10	3	the Api Bank score
Arc E	10	3	the Arc E score
Cnmo 2024	10	3	the Cnmo 2024 score
Frontiermath Tier 4 V2	10	3	the Frontiermath Tier 4 V2 score
Gdpval Mm	10	3	the Gdpval Mm score
Gorilla Benchmark Api Bench	10	3	the Gorilla Benchmark Api Bench score
Medchembench	10	3	the Medchembench score
Mls Bench Lite	10	3	the Mls Bench Lite score
Mmlu Cot	10	3	the Mmlu Cot score
Mmmuval	10	3	the Mmmuval score
Posttrainbench Lite	10	3	the Posttrainbench Lite score
Spreadsheetbench V1	10	3	the Spreadsheetbench V1 score
Frontier Science Research	10	2	the Frontier Science Research score
Humaneval Mul	10	2	the Humaneval Mul score
Kimi Code Bench V2	10	2	the Kimi Code Bench V2 score
Mbpp Evalplus	10	2	the Mbpp Evalplus score
Mimo Coding Bench	10	2	the Mimo Coding Bench score
Natural Questions	10	2	the Natural Questions score
Onemillion Bench	10	2	the Onemillion Bench score
Qwen Swe Bench	10	2	the Qwen Swe Bench score
Qwenwebbench	10	2	the Qwenwebbench score
Qwenworldbench	10	2	the Qwenworldbench score
Swe Atlas Codebase Qna	10	2	the Swe Atlas Codebase Qna score
Swe Bench Verified Agentic Coding	10	2	the Swe Bench Verified Agentic Coding score
Vibe Pro	10	2	the Vibe Pro score
Agent Startup Bench	10	1	the Agent Startup Bench score
Androidbench	10	1	the Androidbench score
Apex Swe	10	1	the Apex Swe score
Artifacts Bench	10	1	the Artifacts Bench score
Automationbench Aa	10	1	the Automationbench Aa score
Bigcodebench Hard	10	1	the Bigcodebench Hard score
Biolp Bench	10	1	the Biolp Bench score
Charxiv Reasoning	10	1	the Charxiv Reasoning score
Cl Bench Life	10	1	the Cl Bench Life score
Finance Agent V1 1	10	1	the Finance Agent V1 1 score
Frontiercode Diamond	10	1	the Frontiercode Diamond score
Gdpval Aa V2 Elo	10	1	the Gdpval Aa V2 Elo score
Gdpval Rubrics	10	1	the Gdpval Rubrics score
Hle Full	10	1	the Hle Full score
Hr Bench 4k	10	1	the Hr Bench 4k score
Imo 2025	10	1	the Imo 2025 score
Infinitebench En Mc	10	1	the Infinitebench En Mc score
Kernel Bench L3	10	1	the Kernel Bench L3 score
Kimi Claw 24 7 Bench	10	1	the Kimi Claw 24 7 Bench score
Livemathematicianbench	10	1	the Livemathematicianbench score
Mcp Universe	10	1	the Mcp Universe score
Measurebench	10	1	the Measurebench score
Miabench	10	1	the Miabench score
Mm Clawbench	10	1	the Mm Clawbench score
Mmsibench	10	1	the Mmsibench score
Next.js Evals	10	1	the Next.js Evals score
Nih Multi Needle	10	1	the Nih Multi Needle score
Openai Mmlu	10	1	the Openai Mmlu score
Ovbench	10	1	the Ovbench score
Plawbench	10	1	the Plawbench score
Prbench Finance	10	1	the Prbench Finance score
Seccodebench	10	1	the Seccodebench score
Spreadsheetbench 2	10	1	the Spreadsheetbench 2 score
Swe Atlas Test Writing	10	1	the Swe Atlas Test Writing score
Swe Mm	10	1	the Swe Mm score
Swe Review	10	1	the Swe Review score
Toolathlon Verified	10	1	the Toolathlon Verified score
Vibe V2	10	1	the Vibe V2 score
Vladbench	10	1	the Vladbench score
Webarena Verified	10	1	the Webarena Verified score
Xdailybench	10	1	the Xdailybench score
MATH	—	86	the MATH score
Arena Code Elo	—	84	the Arena Code Elo score
CritPt	—	84	the CritPt score
Tax Eval v2	—	84	the Tax Eval v2 score
CorpFin v2	—	80	the CorpFin v2 score
MGSM	—	61	the MGSM score
MedCode	—	60	the MedCode score
MedScribe	—	60	the MedScribe score
SAGE	—	55	the SAGE score
CharXiv-R	—	47	CharXiv: can the model read scientific charts.
MedQA	—	47	the MedQA score
IOI	—	45	the IOI score
Alder Polyglot	—	43	the Alder Polyglot score
Code Migration	—	43	the Code Migration score
BFCL	—	42	the BFCL score
FinanceAgent v2	—	42	FinanceAgent: financial analysis tasks, not a generic chat vibe.
EMB	—	40	the EMB score
FinanceAgent v1.1	—	39	FinanceAgent: financial analysis tasks, not a generic chat vibe.
Arena Vision Elo	—	37	the Arena Vision Elo score
Lech Mazur Writing	—	27	the Lech Mazur writing eval, which scores prose quality instead of multiple-choice knowledge.
Ai2d	—	25	the Ai2d score
DROP	—	24	the DROP score
INCLUDE	—	24	the INCLUDE score
Arena Search Elo	—	23	the Arena Search Elo score
MathVision	—	22	the MathVision score
ERQA	—	21	the ERQA score
Hmmt25	—	21	the Hmmt25 score
Multichallenge	—	20	the Multichallenge score
Polymath	—	19	the Polymath score
Docvqa	—	17	the Docvqa score
Graphwalks Bfs 128k	—	17	the Graphwalks Bfs 128k score
Chartqa	—	16	the Chartqa score
FrontierCode	—	16	the FrontierCode score
FrontierCode 1.1	—	16	the FrontierCode 1.1 score
Frontierswe	—	16	the Frontierswe score
Nl2repo	—	16	the Nl2repo score
Omnidocbench 1 5	—	16	the Omnidocbench 1 5 score
Wmt24	—	16	the Wmt24 score
Mbpp	—	15	the Mbpp score
Mmstar	—	15	the Mmstar score
Odinw	—	15	the Odinw score
Charxiv D	—	13	the Charxiv D score
Codeforces	—	13	the Codeforces score
Graphwalks Parents 128k	—	13	the Graphwalks Parents 128k score
Cybergym	—	12	the Cybergym score
Blink	—	11	the Blink score
Hmmt Feb 26	—	11	the Hmmt Feb 26 score
Simplevqa	—	11	the Simplevqa score
Global Piqa	—	10	the Global Piqa score
Infovqatest	—	10	the Infovqatest score
Ocrbench V2 En	—	10	the Ocrbench V2 En score
Widesearch	—	10	the Widesearch score
Aider Polyglot Edit	—	9	the Aider Polyglot Edit score
Babyvision	—	9	the Babyvision score
Charadessta	—	9	the Charadessta score
Collie	—	9	the Collie score
Docvqatest	—	9	the Docvqatest score
Hiddenmath	—	9	the Hiddenmath score
Mlvu	—	9	the Mlvu score
Ocrbench V2 Zh	—	9	the Ocrbench V2 Zh score
Truthfulqa	—	9	the Truthfulqa score
Maxife	—	8	the Maxife score
Mcp Mark	—	8	the Mcp Mark score
Mlvu M	—	8	the Mlvu M score
MMMU-Pro (With Tools)	—	8	the MMMU-Pro (With Tools) score
Arena Agent Elo	—	7	the Arena Agent Elo score
Artificial Analysis Index	—	7	the Artificial Analysis Index score
Csimpleqa	—	7	the Csimpleqa score
Deepsearchqa	—	7	the Deepsearchqa score
Egoschema	—	7	the Egoschema score
Natural2code	—	7	the Natural2code score
Openai Mrcr 2 Needle 128k	—	7	the Openai Mrcr 2 Needle 128k score
Refcoco Avg	—	7	the Refcoco Avg score
Textvqa	—	7	the Textvqa score
V Star	—	7	the V Star score
Videomme W O Sub	—	7	the Videomme W O Sub score
Videomme W Sub	—	7	the Videomme W Sub score
Zebralogic	—	7	the Zebralogic score
Bird Sql Dev	—	6	the Bird Sql Dev score
Dynamath	—	6	the Dynamath score
GRIND	—	6	the GRIND score
Tau3 Banking	—	6	the Tau3 Banking score
Androidworld Sr	—	5	the Androidworld Sr score
Browsecomp Long 128k	—	5	the Browsecomp Long 128k score
Medxpertqa	—	5	the Medxpertqa score
Piqa	—	5	the Piqa score
Swe Lancer Ic Diamond Subset	—	5	the Swe Lancer Ic Diamond Subset score
Zerobench Sub	—	5	the Zerobench Sub score
Aa Index	—	4	the Aa Index score
Boolq	—	4	the Boolq score
Eclektic	—	4	the Eclektic score
Exploitgym	—	4	the Exploitgym score
Fleurs	—	4	the Fleurs score
Graphwalks Parents (0K-128K)	—	4	the Graphwalks Parents (0K-128K) score
Hypersim	—	4	the Hypersim score
Infovqa	—	4	the Infovqa score
Lingoqa	—	4	the Lingoqa score
MMMU-Pro (No Tools)	—	4	the MMMU-Pro (No Tools) score
Mmmu Validation	—	4	the Mmmu Validation score
Mmvu	—	4	the Mmvu score
Nexus	—	4	the Nexus score
OmniDocBench NED	—	4	the OmniDocBench NED score
Openai Mrcr 2 Needle 1m	—	4	the Openai Mrcr 2 Needle 1m score
Openbookqa	—	4	the Openbookqa score
ScienceQA	—	4	the ScienceQA score
Squality	—	4	the Squality score
Sunrgbd	—	4	the Sunrgbd score
Vlmsareblind	—	4	the Vlmsareblind score
Wmt23	—	4	the Wmt23 score
Worldvqa	—	4	the Worldvqa score
Aider	—	3	the Aider score
Amc 2022 23	—	3	the Amc 2022 23 score
Benchcad With Python Tool	—	3	the Benchcad With Python Tool score
Capture The Flag Challenges	—	3	the Capture The Flag Challenges score
Claw Eval Mm	—	3	the Claw Eval Mm score
Corpusqa 1m	—	3	the Corpusqa 1m score
Crag	—	3	the Crag score
Cybersecurity Ctfs	—	3	the Cybersecurity Ctfs score
Figqa	—	3	the Figqa score
Finqa	—	3	the Finqa score
Fullstackbench En	—	3	the Fullstackbench En score
Fullstackbench Zh	—	3	the Fullstackbench Zh score
Graphwalks Bfs 1m	—	3	the Graphwalks Bfs 1m score
Humanitys Last Exam With Tools Text Only	—	3	the Humanitys Last Exam With Tools Text Only score
Kernelgen 1p	—	3	the Kernelgen 1p score
Management Consulting Tasks	—	3	the Management Consulting Tasks score
Math Cot	—	3	the Math Cot score
Mm Mind2web	—	3	the Mm Mind2web score
Mrcr V2 8 Needle 512k 1m	—	3	the Mrcr V2 8 Needle 512k 1m score
Multilingual Mgsm Cot	—	3	the Multilingual Mgsm Cot score
Multipl E Humaneval	—	3	the Multipl E Humaneval score
Multipl E Mbpp	—	3	the Multipl E Mbpp score
Nanogpt	—	3	the Nanogpt score
Nuscene	—	3	the Nuscene score
OfficeQA	—	3	the OfficeQA score
Omniscience	—	3	the Omniscience score
Openai Connectors	—	3	the Openai Connectors score
Openai Search Function Calling	—	3	the Openai Search Function Calling score
Pmc Vqa	—	3	the Pmc Vqa score
Rsi Index	—	3	the Rsi Index score
Ruler	—	3	the Ruler score
Slakevqa	—	3	the Slakevqa score
Social Iqa	—	3	the Social Iqa score
Translation En Set1 Comet22	—	3	the Translation En Set1 Comet22 score
Translation En Set1 Spbleu	—	3	the Translation En Set1 Spbleu score
Translation Set1 En Comet22	—	3	the Translation Set1 En Comet22 score
Translation Set1 En Spbleu	—	3	the Translation Set1 En Spbleu score
Usamo 2026	—	3	the Usamo 2026 score
Vision2web	—	3	the Vision2web score
Vqav2	—	3	the Vqav2 score
Vqav2 Val	—	3	the Vqav2 Val score
Xstest	—	3	the Xstest score
Alpacaeval 2 0	—	2	the Alpacaeval 2 0 score
Apex	—	2	the Apex score
Arxivmath	—	2	the Arxivmath score
Autologi	—	2	the Autologi score
Beyond Aime	—	2	the Beyond Aime score
Bfcl V2	—	2	the Bfcl V2 score
Bfcl V3 Multiturn	—	2	the Bfcl V3 Multiturn score
Browsecomp Long 256k	—	2	the Browsecomp Long 256k score
Charxiv Reasoning No Tools	—	2	the Charxiv Reasoning No Tools score
Charxiv Reasoning With Tools	—	2	the Charxiv Reasoning With Tools score
Cluewsc	—	2	the Cluewsc score
Covost2	—	2	the Covost2 score
Design2code	—	2	the Design2code score
Dsbench Fullstack	—	2	the Dsbench Fullstack score
Dsbench Hard	—	2	the Dsbench Hard score
Factscore	—	2	the Factscore score
Frames	—	2	the Frames score
Frontiercode Diamond Xhigh	—	2	the Frontiercode Diamond Xhigh score
Frontierscience Olympiad	—	2	the Frontierscience Olympiad score
Frontierscience Research	—	2	the Frontierscience Research score
Functionalmath	—	2	the Functionalmath score
Gdm Mrcr V2 8needle 128k Average	—	2	the Gdm Mrcr V2 8needle 128k Average score
Gdm Mrcr V2 8needle 1m Pointwise	—	2	the Gdm Mrcr V2 8needle 1m Pointwise score
Gdp Pdf No Tools	—	2	the Gdp Pdf No Tools score
Graphwalks Parents (256K-1M)	—	2	the Graphwalks Parents (256K-1M) score
Groundui 1k	—	2	the Groundui 1k score
Gsm 8k Cot	—	2	the Gsm 8k Cot score
Horizonmath	—	2	the Horizonmath score
Humanitys Last Exam No Tools Text Only	—	2	the Humanitys Last Exam No Tools Text Only score
If	—	2	the If score
Mmsearch Plus	—	2	the Mmsearch Plus score
Mobileworld	—	2	the Mobileworld score
Multilf	—	2	the Multilf score
Musr	—	2	the Musr score
OpenAI MRCR v2 8-needle (128K-256K)	—	2	the OpenAI MRCR v2 8-needle (128K-256K) score
OpenAI MRCR v2 8-needle (128K-512K)	—	2	the OpenAI MRCR v2 8-needle (128K-512K) score
OpenAI MRCR v2 8-needle (256K-1M)	—	2	the OpenAI MRCR v2 8-needle (256K-1M) score
OpenAI MRCR v2 8-needle (32K-128K)	—	2	the OpenAI MRCR v2 8-needle (32K-128K) score
OpenAI MRCR v2 8-needle (4K-8K)	—	2	the OpenAI MRCR v2 8-needle (4K-8K) score
OpenAI MRCR v2 8-needle (512K-1M)	—	2	the OpenAI MRCR v2 8-needle (512K-1M) score
OpenAI MRCR v2 8-needle (64K-128K)	—	2	the OpenAI MRCR v2 8-needle (64K-128K) score
Perceptionbench	—	2	the Perceptionbench score
Physicsfinals	—	2	the Physicsfinals score
Polymath En	—	2	the Polymath En score
Qwen Svg	—	2	the Qwen Svg score
Superchem	—	2	the Superchem score
Swe Bench Verified Agentless	—	2	the Swe Bench Verified Agentless score
Tydiqa	—	2	the Tydiqa score
Usamo25	—	2	the Usamo25 score
Vatex	—	2	the Vatex score
Videoholmes	—	2	the Videoholmes score
Visfactor	—	2	the Visfactor score
Visulogic	—	2	the Visulogic score
Aa Briefcase Elo	—	1	the Aa Briefcase Elo score
Aa Omniscience Index	—	1	the Aa Omniscience Index score
Activitynet	—	1	the Activitynet score
Advanced Cybersecurity Completion Rate	—	1	the Advanced Cybersecurity Completion Rate score
Aethercode	—	1	the Aethercode score
Ai2 Reasoning Challenge Arc	—	1	the Ai2 Reasoning Challenge Arc score
Aime	—	1	the Aime score
Aitz Em	—	1	the Aitz Em score
Android Control High Em	—	1	the Android Control High Em score
Android Control Low Em	—	1	the Android Control Low Em score
Arc	—	1	the Arc score
Arcagi2	—	1	the Arcagi2 score
Arkitscenes	—	1	the Arkitscenes score
Attaq	—	1	the Attaq score
Babyvision With Python	—	1	the Babyvision With Python score
Bc Vl	—	1	the Bc Vl score
Beam 128k	—	1	the Beam 128k score
Bigcodebench Full	—	1	the Bigcodebench Full score
Biomysterybench Hard	—	1	the Biomysterybench Hard score
Biomysterybench Human Solved	—	1	the Biomysterybench Human Solved score
Cbnsl	—	1	the Cbnsl score
Cc Bench V2 Backend	—	1	the Cc Bench V2 Backend score
Cc Bench V2 Frontend	—	1	the Cc Bench V2 Frontend score
Cc Bench V2 Repo	—	1	the Cc Bench V2 Repo score
Chartmuseum	—	1	the Chartmuseum score
Chartqapro	—	1	the Chartqapro score
Charxiv Rq	—	1	the Charxiv Rq score
Charxiv Rq With Python	—	1	the Charxiv Rq With Python score
Ci Memories Coverage	—	1	the Ci Memories Coverage score
Ci Memories Violation	—	1	the Ci Memories Violation score
Cloningscenarios	—	1	the Cloningscenarios score
Codegolf V2 2	—	1	the Codegolf V2 2 score
Cohere Agentic Question Answering	—	1	the Cohere Agentic Question Answering score
Cohere Data Analysis	—	1	the Cohere Data Analysis score
Cohere Memory Usage Quality	—	1	the Cohere Memory Usage Quality score
Commonsenseqa	—	1	the Commonsenseqa score
Contphy	—	1	the Contphy score
Countqa	—	1	the Countqa score
Creativework	—	1	the Creativework score
Crossvid	—	1	the Crossvid score
Crux O	—	1	the Crux O score
Cursorbench 3 2	—	1	the Cursorbench 3 2 score
Dailyomni	—	1	the Dailyomni score
Deepsearchqa F1	—	1	the Deepsearchqa F1 score
Deepswe 1 0	—	1	the Deepswe 1 0 score
Deepswe 1 0 Pass At 1	—	1	the Deepswe 1 0 Pass At 1 score
Deepswe 1 1 Mini Swe Agent	—	1	the Deepswe 1 1 Mini Swe Agent score
Doubao Multi Turn Bench	—	1	the Doubao Multi Turn Bench score
Draco	—	1	the Draco score
Ds Arena Code	—	1	the Ds Arena Code score
Ds Fim Eval	—	1	the Ds Fim Eval score
Dude	—	1	the Dude score
Emma	—	1	the Emma score
Exploitbench Cap Percent	—	1	the Exploitbench Cap Percent score
Finsearchcomp T2 T3	—	1	the Finsearchcomp T2 T3 score
Finsearchcomp T3	—	1	the Finsearchcomp T3 score
Flame Vlm Code	—	1	the Flame Vlm Code score
French Mmlu	—	1	the French Mmlu score
Frontier Swe Impl	—	1	the Frontier Swe Impl score
Frontiercode Main	—	1	the Frontiercode Main score
Frontiercs	—	1	the Frontiercs score
Gaia2	—	1	the Gaia2 score
Gameworld	—	1	the Gameworld score
Gdp Pdf Mean Criteria No Tools	—	1	the Gdp Pdf Mean Criteria No Tools score
Gdp Pdf Mean Criteria With Tools	—	1	the Gdp Pdf Mean Criteria With Tools score
Gdp Pdf Strict Pass	—	1	the Gdp Pdf Strict Pass score
Govreport	—	1	the Govreport score
Gpqa Biology	—	1	the Gpqa Biology score
Gpqa Chemistry	—	1	the Gpqa Chemistry score
Gpqa Physics	—	1	the Gpqa Physics score
Harvey Lab Aa	—	1	the Harvey Lab Aa score
Hipho	—	1	the Hipho score
Hle Full With Tools	—	1	the Hle Full With Tools score
Hle Verified	—	1	the Hle Verified score
Humaneval Er	—	1	the Humaneval Er score
Image2floorplan	—	1	the Image2floorplan score
Imagemining	—	1	the Imagemining score
Infinitebench En Qa	—	1	the Infinitebench En Qa score
Infographicsqa	—	1	the Infographicsqa score
Intergps	—	1	the Intergps score
Kina	—	1	the Kina score
Livecodebench 01 09	—	1	the Livecodebench 01 09 score
Livesports 3k	—	1	the Livesports 3k score
Loca Bench 256k	—	1	the Loca Bench 256k score
Longfact Concepts	—	1	the Longfact Concepts score
Longfact Objects	—	1	the Longfact Objects score
Lsat	—	1	the Lsat score
Mask	—	1	the Mask score
Mathverse	—	1	the Mathverse score
Mathverse Mini	—	1	the Mathverse Mini score
Mathvision With Python	—	1	the Mathvision With Python score
Mbpp Base Version	—	1	the Mbpp Base Version score
Mbpp Evalplus Base	—	1	the Mbpp Evalplus Base score
Mbpp Pass 1	—	1	the Mbpp Pass 1 score
Mbpp Plus	—	1	the Mbpp Plus score
Medxpertqa Mm	—	1	the Medxpertqa Mm score
Mega Mlqa	—	1	the Mega Mlqa score
Mega Tydi Qa	—	1	the Mega Tydi Qa score
Mega Udpos	—	1	the Mega Udpos score
Mega Xcopa	—	1	the Mega Xcopa score
Mega Xstorycloze	—	1	the Mega Xstorycloze score
Mewc	—	1	the Mewc score
Minerva	—	1	the Minerva score
Mm If Eval	—	1	the Mm If Eval score
Mmau	—	1	the Mmau score
Mmbc	—	1	the Mmbc score
Mmlongbench 128k	—	1	the Mmlongbench 128k score
Mmlu French	—	1	the Mmlu French score
Mmmu Pro With Python	—	1	the Mmmu Pro With Python score
Mmsearch	—	1	the Mmsearch score
Mmvet	—	1	the Mmvet score
Mobileminiwob Sr	—	1	the Mobileminiwob Sr score
Mrcr 128k 8 Needle	—	1	the Mrcr 128k 8 Needle score
Mrcr 1m Pointwise	—	1	the Mrcr 1m Pointwise score
Msqa	—	1	the Msqa score
Mt Aime 2025	—	1	the Mt Aime 2025 score
Next.js Evals + AGENTS.md	—	1	the Next.js Evals + AGENTS.md score
Objectron	—	1	the Objectron score
Officeqa Pro Vision	—	1	the Officeqa Pro Vision score
Ojbench Cpp	—	1	the Ojbench Cpp score
Omniscience Non Hallucination Rate	—	1	the Omniscience Non Hallucination Rate score
Open Rewrite	—	1	the Open Rewrite score
Openai Mrcr 2 Needle 256k	—	1	the Openai Mrcr 2 Needle 256k score
Openrca	—	1	the Openrca score
Osworld Extended	—	1	the Osworld Extended score
Osworld Screenshot Only	—	1	the Osworld Screenshot Only score
Perceptiontest	—	1	the Perceptiontest score
Phibench	—	1	the Phibench score
Pope	—	1	the Pope score
Popqa	—	1	the Popqa score
Prbench Legal	—	1	the Prbench Legal score
Protocolqa	—	1	the Protocolqa score
Qasper	—	1	the Qasper score
Qmsum	—	1	the Qmsum score
Qvhighlights	—	1	the Qvhighlights score
Qwen Qoder Bench	—	1	the Qwen Qoder Bench score
Qwen React Bench	—	1	the Qwen React Bench score
Realkie Fcc	—	1	the Realkie Fcc score
Recreationbench	—	1	the Recreationbench score
Repo Env	—	1	the Repo Env score
Repoqa	—	1	the Repoqa score
Robospatialhome	—	1	the Robospatialhome score
Sat Math	—	1	the Sat Math score
Scienceqa Visual	—	1	the Scienceqa Visual score
Seedclawbench	—	1	the Seedclawbench score
Sifo	—	1	the Sifo score
Sifo Multiturn	—	1	the Sifo Multiturn score
Siren Agentdojo Attack Success	—	1	the Siren Agentdojo Attack Success score
Siren Agentdojo Utility	—	1	the Siren Agentdojo Utility score
Spider	—	1	the Spider score
Summscreenfd	—	1	the Summscreenfd score
Superglue	—	1	the Superglue score
Surds	—	1	the Surds score
Swe Bench Pro Resolve Rate	—	1	the Swe Bench Pro Resolve Rate score
Swe Bench Verified Multiple Attempts	—	1	the Swe Bench Verified Multiple Attempts score
Swe Marathon Resolution Rate	—	1	the Swe Marathon Resolution Rate score
Tau3 Airline	—	1	the Tau3 Airline score
Tau3 Retail	—	1	the Tau3 Retail score
Tau3 Telecom	—	1	the Tau3 Telecom score
Tempcompass	—	1	the Tempcompass score
Terminus	—	1	the Terminus score
Theoremqa	—	1	the Theoremqa score
Tldr9 Test	—	1	the Tldr9 Test score
Tomato	—	1	the Tomato score
Trae Code Gen	—	1	the Trae Code Gen score
Trae Error Fix	—	1	the Trae Error Fix score
Uniform Bar Exam	—	1	the Uniform Bar Exam score
Vct	—	1	the Vct score
Vibe	—	1	the Vibe score
Vibe Android	—	1	the Vibe Android score
Vibe Backend	—	1	the Vibe Backend score
Vibe Ios	—	1	the Vibe Ios score
Vibe Simulation	—	1	the Vibe Simulation score
Vibe Web	—	1	the Vibe Web score
Video Mme Long No Subtitles	—	1	the Video Mme Long No Subtitles score
Videosimpleqa	—	1	the Videosimpleqa score
Vlmsarebiased	—	1	the Vlmsarebiased score
Voicebench Avg	—	1	the Voicebench Avg score
Vqav2 Test	—	1	the Vqav2 Test score
We Math	—	1	the We Math score
Webvoyager	—	1	the Webvoyager score
Wmdp	—	1	the Wmdp score
Worldvqa Forceanswer	—	1	the Worldvqa Forceanswer score
Zerobench Main Pass At 5	—	1	the Zerobench Main Pass At 5 score
Zerobench Main With Python Pass At 5	—	1	the Zerobench Main With Python Pass At 5 score