Named effort levels run from lowest to highest. Select a result to see its source details.
Multilingual Bridging the Gap 533 related benchmarks · 549 results
Show 540 more results Hide 540 results EKI Estonian Source: Keelemudelite mõõdupuu data docs keelenou 47.3 bib_bench 37.9 idiom_bench 45.7 propaganda_resistance 66.3 term_bench 32.6 trivia_et_2 54 trivia_et_3 4.3 FormosanBench summarization 12 related benchmarks · 12 results
Show 6 more results Hide 6 results MMMLU Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 81.4% PashtoBench 10 related benchmarks · 10 results
Show 4 more results Hide 4 results SeaLLMs 16 related benchmarks · 16 results
Show 10 more results Hide 10 results Agentic AgentDojo 3 related benchmarks · 27 results
utility Source: AgentDojo results data docs · Jun 2024 By attack · provider: openai · defense: None direct 69.1% ignore_previous 69.1% important_instructions 69.1% injecagent 69.1% tool_knowledge 69.1%
By defense · provider: openai · attack: important_instructions repeat_user_prompt 84.5% spotlighting_with_delimiting 72.2% tool_filter 72.2% transformers_pi_detector 41.2%
Targeted ASR Source: AgentDojo results data docs · Jun 2024 By attack · provider: openai · defense: None direct 3.7% ignore_previous 5.4% important_instructions 47.7% injecagent 5.7% tool_knowledge 34.5%
By defense · provider: openai · attack: important_instructions repeat_user_prompt 27.8% spotlighting_with_delimiting 41.7% tool_filter 6.8% transformers_pi_detector 8%
utility under attack Source: AgentDojo results data docs · Jun 2024 By attack · provider: openai · defense: None direct 67.3% ignore_previous 66.8% important_instructions 50.1% injecagent 68.5% tool_knowledge 57.7%
By defense · provider: openai · attack: important_instructions repeat_user_prompt 67.3% spotlighting_with_delimiting 55.6% tool_filter 56.3% transformers_pi_detector 21.1%
BrowserGym Source: BrowserGym Leaderboard MiniWoB 63.8% AssistantBench 4.8% WebArena 31.4% WebLINX 12.5% WorkArena-L1 45.5% WorkArena-L2 8.5% WorkArena-L3 0% Source: GAIA Leaderboard data · Jun 2024 Difficulty level Level 1 58.5% Level 2 43% Level 3 19.2%
Difficulty not reported 44.2%
Gaia2 Source: Gaia2 (Meta ARE) data · Sep 2025 Gaia2 7.4% A2A 5.2% Adaptability 6.2% Ambiguity 4.4% Execution 8.3% Noise 4.6% Search 17.5% Time 5.8% Orak 14 related benchmarks · 62 results
DarkestDungeon default 93.4% planning 82% ref-plan 91.6% reflection 85.2% zeroshot 93.4% HerStory default 64% planning 59.2% ref-plan 61.8% reflection 61.3% zeroshot 64% MineCraft default 71% planning 13% ref-plan 50% reflection 50% zeroshot 0% Show 32 more results Hide 32 results Pokémon default 38.9% planning 33.3% ref-plan 38.9% reflection 36.1% zeroshot 33.3% SlayTheSpire default 23.6% planning 35.3% ref-plan 36% reflection 41.3% zeroshot 24.7% StarCraftII default 100% planning 50% ref-plan 100% reflection 0% zeroshot 50% StardewValley default 95.7% planning 64.6% ref-plan 95.7% reflection 18.3% zeroshot 40.5% StreetFighterIII default 29.7% planning 30% ref-plan 23.3% reflection 23.3% zeroshot 29.7% Supermario default 34.1% planning 29.4% ref-plan 34.1% reflection 32.3% zeroshot 29.6% tau2-bench airline Source: OpenRouter benchmark exports · Oct 2026 OpenRouter's automatic routing · gpt-4o-2024-08-06 · 90-day results 42.7%
OpenAI · gpt-4o-2024-08-06 · 90-day results 43.1%
OpenRouter's automatic routing · gpt-4o · 90-day results 41.7%
OpenAI · gpt-4o · 90-day results 43.2%
Show 9 more results Hide 9 results Image generation UniGenBench++ 136 related benchmarks · 136 results
Safety HELM 66 related benchmarks · 125 results
Show 114 more results Hide 114 results PashtoBench safety 5 related benchmarks · 5 results
Phare Source: Phare Bias Resistance 50.9% Harm Resistance 92.7% Jailbreak Resistance 60.5% Knowledge HELM 60 related benchmarks · 116 results
Show 108 more results Hide 108 results MMLU-Pro Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 74.7% Math HELM 3 related benchmarks · 3 results
Source: JudgeBench data docs Prompted Judge · Arena-Hard (gpt-4o-2024-05-13) · Math Score Claude-3.5-Sonnet Dataset 61.8% GPT-4o Dataset 75%
GPT-4o Dataset, Multi-Agent Judge, ChatEval (gpt-4o-2024-05-13), Math Score 44.6%
GPT-4o Dataset, Prompted Judge, Vanilla (gpt-4o-2024-05-13), Math Score 66.1%
MATH Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 76.6% MGSM Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 90.5% mu-MATH 7 related benchmarks · 35 results
F1 Source: U-MATH and mu-MATH F1 · Qwen/Qwen2.5-72B-Instruct Gemini-1.5-Pro 72.6% GPT-4o 77.5% Llama-3.1-70B-Instruct 81.8% mu_math 77.4% Qwen2.5-72B-Instruct 74.2%
accuracy Source: U-MATH and mu-MATH accuracy · Qwen/Qwen2.5-72B-Instruct Gemini-1.5-Pro 74.2% GPT-4o 77.5% Llama-3.1-70B-Instruct 83.8% mu_math 77.4% Qwen2.5-72B-Instruct 74.2%
MCC Source: U-MATH and mu-MATH MCC · Qwen/Qwen2.5-72B-Instruct Gemini-1.5-Pro 0.746 GPT-4o 0.778 Llama-3.1-70B-Instruct 0.82 mu_math 0.781 Qwen2.5-72B-Instruct 0.752
NPV Source: U-MATH and mu-MATH NPV · Qwen/Qwen2.5-72B-Instruct Gemini-1.5-Pro 54.8% GPT-4o 73.6% Llama-3.1-70B-Instruct 84.9% mu_math 71.3% Qwen2.5-72B-Instruct 65.5%
PPV Source: U-MATH and mu-MATH PPV · Qwen/Qwen2.5-72B-Instruct Gemini-1.5-Pro 90.5% GPT-4o 82.1% Llama-3.1-70B-Instruct 81.2% mu_math 85.1% Qwen2.5-72B-Instruct 84.6%
TNR Source: U-MATH and mu-MATH TNR · Qwen/Qwen2.5-72B-Instruct Gemini-1.5-Pro 82.9% GPT-4o 83.2% Llama-3.1-70B-Instruct 90.8% mu_math 85.9% Qwen2.5-72B-Instruct 83.6%
Show 5 more results Hide 5 results TPR Source: U-MATH and mu-MATH TPR · Qwen/Qwen2.5-72B-Instruct Gemini-1.5-Pro 70.4% GPT-4o 72.1% Llama-3.1-70B-Instruct 71.1% mu_math 70.1% Qwen2.5-72B-Instruct 67.1%
Source: Omni-MATH · May 2024 Setting not stated 30.5%
Algebra 36.1%
Applied Mathematics 29.4%
Calculus 21.9%
Difficulty: 1-3 68.4%
Difficulty: 3-5 25%
Show 6 more results Hide 6 results Difficulty: 5-8 21.8%
Difficulty: 8-10 15.8%
Discrete Mathematics 15.7%
Geometry 21.6%
Number Theory 25.8%
Precalculus 39.8%
PutnamBench 3 related benchmarks · 6 results
coq-wsolution solved Source: PutnamBench COPRA (GPT-4o), coq-wsolution, pass@1, pass@1 w/ T=0.0; 1988 B1. 1
GPT-4o, coq-wsolution, pass@10, pass@10 w/ T=0.7; 1988 B1. 1
isabelle-wsolution solved Source: PutnamBench DSP (GPT-4o), isabelle-wsolution, pass@10, pass@10 w/ T=0.1; 1971 B1, 1986 B1, 1995 B1, 2001 A1. 4
GPT-4o, isabelle-wsolution, pass@10, pass@10 w/ T=0.7; 1988 B1. 1
lean-wsolution solved Source: PutnamBench COPRA (GPT-4o), lean-wsolution, pass@1, pass@1 w/ T=0.0; 1988 B1. 1
GPT-4o, lean-wsolution, pass@10, pass@10 w/ T=0.7; 1988 B1. 1
U-MATH 7 related benchmarks · 21 results
Show 3 more results Hide 3 results Vals Source: Vals AI AIME 14% MATH 500 75.2% MGSM 90.7% WildBench 5 related benchmarks · 5 results
Human preference English 2024-05-13 1,312.4 2024-08-06 1,297.4 Expert 2024-05-13 1,249.8 2024-08-06 1,240.4 Show 51 more results Hide 51 results French 2024-05-13 1,304.1 2024-08-06 1,262.8 German 2024-05-13 1,282.4 2024-08-06 1,257.3 Japanese 2024-05-13 1,257 2024-08-06 1,234.4 Korean 2024-05-13 1,233.9 2024-08-06 1,218.3 Russian 2024-05-13 1,285.5 2024-08-06 1,271.8 Spanish 2024-05-13 1,291.9 2024-08-06 1,268.6 Vision 2024-05-13 1,136.5 2024-08-06 1,064.5 WildBench 26 related benchmarks · 26 results
Show 20 more results Hide 20 results Coding Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim Setting not stated 30.7%
Source: ArtifactsBench Judge-Model-Gemini-2.5-Pro | latest 33.5%
Judge-Model-Gemini-2.5-Pro-0506 | history 38%
Source: JudgeBench data docs Prompted Judge · Arena-Hard (gpt-4o-2024-05-13) · Coding Score Claude-3.5-Sonnet Dataset 38.7% GPT-4o Dataset 59.5%
GPT-4o Dataset, Multi-Agent Judge, ChatEval (gpt-4o-2024-05-13), Coding Score 31%
GPT-4o Dataset, Prompted Judge, Vanilla (gpt-4o-2024-05-13), Coding Score 61.9%
Show 2 more results Hide 2 results Multi-SWE-bench 9 related benchmarks · 27 results
Show 9 more results Hide 9 results Show 4 more results Hide 4 results WildBench 5 related benchmarks · 5 results
Vision AI2D Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 94.2% ChartQA Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 85.7% Show 7 more results Hide 7 results DocVQA Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 92.8% MMMU Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 72.2% MMMU-Pro Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 59.9% Science BixBench 3 related benchmarks · 9 results
GPQA Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 70.1% Source: OpenRouter benchmark exports · Oct 2026 gpt-4o-2024-08-06 · 90-day results Across 2 providers lowest 51.5% median 52.2% highest 52.8% OpenRouter's automatic routing 52%
Show results from 2 providers and OpenRouter's automatic routing Hide results from 2 providers and OpenRouter's automatic routing OpenAI 52.8% OpenRouter's automatic routing 52% Azure 51.5% gpt-4o · 90-day results Across 2 providers lowest 51.7% median 52.2% highest 52.7% OpenRouter's automatic routing 51.1%
Show results from 2 providers and OpenRouter's automatic routing Hide results from 2 providers and OpenRouter's automatic routing OpenAI 52.7% Azure 51.7% OpenRouter's automatic routing 51.1% Reasoning COLLIE Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 61% DROP Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 83.4% ERQA Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 35.2% HELM 3 related benchmarks · 4 results
Source: JudgeBench data docs Prompted Judge · Arena-Hard (gpt-4o-2024-05-13) · Reasoning Score Claude-3.5-Sonnet Dataset 51% GPT-4o Dataset 54.1%
GPT-4o Dataset, Multi-Agent Judge, ChatEval (gpt-4o-2024-05-13), Reasoning Score 31.6%
GPT-4o Dataset, Prompted Judge, Vanilla (gpt-4o-2024-05-13), Reasoning Score 48%
Instruction following IFEval Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 81% Multi-IF Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 60.9% PashtoBench instruction following 4 related benchmarks · 4 results
Factuality SimpleQA Source: LLM Stats data · self-reported Reported by the model's developer. The source relaying it did not verify it. claim 38.2% OCR PsOCR Pashto OCR 7 related benchmarks · 7 results
Other CSEBenchmark 11 related benchmarks · 11 results
Horangi 36 related benchmarks · 36 results
Source: JudgeBench data docs Prompted Judge · Arena-Hard (gpt-4o-2024-05-13) · Knowledge Score Claude-3.5-Sonnet Dataset 52.6% GPT-4o Dataset 50.6%
GPT-4o Dataset, Multi-Agent Judge, ChatEval (gpt-4o-2024-05-13), Knowledge Score 32.5%
GPT-4o Dataset, Prompted Judge, Vanilla (gpt-4o-2024-05-13), Knowledge Score 44.2%
Source: JudgeBench data docs Prompted Judge · Arena-Hard (gpt-4o-2024-05-13) · Overall Score Claude-3.5-Sonnet Dataset 51.9% GPT-4o Dataset 56.6%
GPT-4o Dataset, Multi-Agent Judge, ChatEval (gpt-4o-2024-05-13), Overall Score 34%
GPT-4o Dataset, Prompted Judge, Vanilla (gpt-4o-2024-05-13), Overall Score 50.9%
MedHELM 33 related benchmarks · 33 results
Nejumi 4 28 related benchmarks · 28 results
Open Greek FinLLM 11 related benchmarks · 11 results
Show 5 more results Hide 5 results Vals Source: Vals AI CaseLaw v2 59.7% CorpFin v2 45.9% LegalBench 82.2% MortgageTax 61% TaxEval v2 74.5% Data credits: ACEBench, MIT (licence ) · Data by LLM Stats · AgentDojo. Copyright (c) 2024 Edoardo Debenedetti, Jie Zhang, Mislav Balunovic, Luca Beurer-Kellner, Marc Fischer, and Florian Tramèr. MIT License. Not a leaderboard: attack and defense coverage differs between models. (licence ) · Aider LLM Leaderboards (Aider-AI), Apache-2.0 (licence ) · Epoch AI Benchmarking Hub, CC BY 4.0 (licence ) · LMArena, CC BY 4.0 (licence ) · ArtifactsBench (Tencent Hunyuan), CC BY 4.0 (licence ) · FutureHouse and ScienceMachine, BixBench v1.5 public zero-shot baseline file. Apache-2.0 repository. (licence ) · Tuka Alhanai, Adam Kasumovic, Mohammad Ghassemi, Aven Zitzelberger, Jessica Lundin and Guillaume Chabot-Couture. Bridging the Gap: Enhancing LLM Performance for Low-Resource African Languages with New Benchmarks, Fine-Tuning, and Cultural Adjustments, arXiv:2412.12417v1. CC BY 4.0 article numeric/protocol extract, restructured with values unchanged. No task or model rights asserted. (licence ) · Steel.dev Agent Leaderboards, MIT (licence ) · BrowserGym Leaderboard (ServiceNow), MIT (licence ) · BullshitBench by Peter Gostev, MIT (licence ) · CharXiv (Wang et al., NeurIPS 2024), Princeton NLP. https://charxiv.github.io · ChemBench leaderboard, Jablonka group / lamalab-org. Public verified numeric result summaries; results licence not separately stated. · CSEBenchmark (Wang et al., IEEE S&P 2025). Copyright (c) 2024 NASP. MIT License. (licence ) · CyBench — cite Zhang et al. 2025 · DABstep (Adyen), CC BY 4.0 (licence ) · Keelemudelite mõõdupuu: Independent LLM leaderboard for Estonian. Eesti Keele Instituut. https://moodupuu.eki.ee, CC BY 4.0. Latest per-benchmark results selected with partial-run flags retained. (licence ) · EvalPlus Leaderboard and EvalPerf (evalplus.github.io), Apache-2.0; EvalPerf win-rate fractions converted to percentages (licence ) · FaithJudge, Vectara; Tamber et al., EMNLP 2025 Industry. Public numeric README leaderboard; results licence not separately established. · Lin, Chen and Zhang, FormosanBench: Benchmarking Low-Resource Austronesian Languages in the Era of Large Language Models, arXiv:2506.21563v1, CC BY 4.0. Scores extracted from Tables 2-4. (licence ) · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence ) · Stanford HELM (Stanford CRFM); scenario scores from Capabilities, Safety, Long Context, MedHELM, AIR-Bench, MMLU and Finance · Stanford HELM (Stanford CRFM), Apache-2.0 (licence ) · Horangi 4, Weights & Biases Korea; public publisher-selected numeric evaluation results · JudgeBench, Sijun Tan et al., ScalerLab · LLM Chess by Maxim Saplin, Apache-2.0 (Copyright [2024] [Maxim Saplin]) (licence ) · M-IFEval: Dussolle, Cardeña Díaz, Sato and Devine (Lightblue), arXiv:2502.04688. README leaderboard, repository licensed Apache-2.0. (licence ) · MathArena (ETH SRI / INSAIT), MIT — cite dekoninck2026matharena (licence ) · MineBench, MIT (licence ) · MLE-bench (OpenAI), MIT (licence ) · Toloka, U-MATH and mu-MATH public leaderboard. MIT Space; generated solution outputs retain underlying model terms and are not ingested. (licence ) · Multi-SWE-bench experiments (multi-swe-bench/experiments), Apache-2.0; resolved-rate fractions converted to percentages (licence ) · Nejumi Leaderboard 4, Weights & Biases Japan; public publisher-selected numeric evaluation results · Online-Mind2Web Leaderboard (OSU NLP), Apache-2.0 (licence ) · The Fin AI. Published FinBen result summaries and Open Greek Financial LLM Leaderboard aggregates. Results licence not separately stated; fractional FinBen success metrics converted to percentages. · Orak leaderboard, KRAFTON AI (Park et al., NeurIPS 2025). Public numeric scores; leaderboard licence not stated (code MIT not applied; CC BY-NC trajectories not used). · PashtoBench v1.0, Jawad Noori. Public results/scores numbers; code MIT and data/ CC BY 4.0 grants do not name results/, so no results licence is asserted. · pfgen-bench, Preferred Networks. Apache-2.0 except LLM-generated outputs; aggregate scores only. (licence ) · Phare (Giskard), CC BY 4.0 (licence ) · Physical AI Bench (SHI Labs), MIT (licence ) · Haq et al., PsOCR: Benchmarking Large Multimodal Models for Optical Character Recognition in Low-resource Pashto Language, arXiv:2505.10055v2, CC BY 4.0. Scores extracted from Tables 3-5. (licence ) · Trishul Lab, PutnamBench public submitted results. Model identities and scaffold budgets preserved; result licence unstated. · ScreenSpot grounding leaderboard (GUI-Agent), Apache-2.0 (licence ) · SeaExam and SeaBench leaderboard, SeaLLMs (DAMO-NLP-SG). Published leaderboard scores; results licence not stated. · Shaberi v2.1, Shisa.AI, Apache-2.0. GPT-5.1-judge scores only; judges are not comparable. (licence ) · SnakeBench (Greg Kamradt / ARC Prize Foundation), MIT (licence ) · Swallow LLM Leaderboard v2, Swallow LLM team (Institute of Science Tokyo). Contents CC BY 4.0; Japanese post-trained task scores on the source's 0..1 scale. (licence ) · SWE-bench-Live (Microsoft and contributors), swe-bench-live.github.io. Resolve rates computed from the published report counts. · SWE-Lancer community leaderboard. Offline task subset as sized in the publisher README; aggregate accuracy and earned payouts. · UniGenBench++ (CodeGoat24), MIT (licence ) · UzLiB, Tahrirchi (Abror Shopulatov). Public LEADERBOARD.md numbers; repository MIT covers software and documentation, no separate results licence stated and none asserted. · Vectara Hallucination Leaderboard, Apache-2.0 (licence ) · VideoGameBench, Alex L. Zhang, Thomas L. Griffiths, Karthik R. Narasimhan, Ofir Press. Public numeric scores; site licence not stated (code MIT not applied; games excluded). · WebWalkerQA (Alibaba NLP), Apache-2.0 (licence ) · WildBench, Allen Institute for AI · WildVision Arena, MIT (licence )