Every AI leaderboard

Leaderboards28Models ranked986

28 leaderboards · 986 models ranked

Every AI leaderboard we track

ChatLMArenaLMArena Text ArenaBlind pairwise votes from millions of people chatting with two anonymous models. Style control on.
  1. 1Gemini 4 Carbon1,526
  2. 2Gemini 4 Argon1,523
  3. 3Muse Wonder1,513
Published Nov 6, 2026Top 10 →ChatScale AIScale SEAL MultiChallengeMulti-turn conversations that test instruction retention, inference memory and self-coherence.
  1. 1GPT-6.1 Sol82.0%
  2. 2Gemini 4 Argon78.7%
  3. 3Muse Spark75.5%
Published Nov 9, 2026Top 10 →ReasoningArtificial AnalysisArtificial Analysis Intelligence IndexA composite of ten independent evaluations run by Artificial Analysis: reasoning, knowledge, math, coding, agents.
  1. 1Gemini 4 Carbon60.0
  2. 2Claude Opus 5.559.1
  3. 3GPT-6 Astra56.9
Published Nov 9, 2026Top 10 →ReasoningScale AI and CAISHumanity's Last Exam2,500 expert-written questions across dozens of subjects, built to be the last closed-ended academic exam.
  1. 1GPT-6 Astra54.8%
  2. 2Claude Fable 5.146.5%
  3. 3Gemini 3.1 Pro Preview46.4%
Published Nov 9, 2026Top 10 →ReasoningEpoch AIGPQA Diamond (Epoch AI runs)198 graduate-level biology, physics and chemistry questions that non-experts cannot look up. Run by Epoch AI.
  1. 1Gemini 4 Argon96.4%
  2. 2GPT-6 Astra95.8%
  3. 3Claude Sonnet 5.595.6%
Published Nov 6, 2026Top 10 →ReasoningEpoch AIEpoch Capabilities IndexOne capability score stitched together from dozens of benchmarks with item response theory.
  1. 1GPT-6 Astra169.1
  2. 2GPT-6.1 Sol168.9
  3. 3Claude Opus 5.5168.4
Published Nov 9, 2026Top 10 →ReasoningARC PrizeARC-AGI-2Abstract visual puzzles that are easy for people and hard for AI. Semi-private set, systems that are plain models only.
  1. 1GPT-6 Astra95.0%
  2. 2GPT-6.1 Sol94.2%
  3. 3Claude Opus 5.593.3%
Published Nov 6, 2026Top 10 →ReasoningLiveBenchLiveBenchContamination-limited questions refreshed regularly: math, reasoning, coding, data analysis, language, instruction following.
  1. 1Claude Fable 5.183.4%
  2. 2Claude Opus 5.583.2%
  3. 3Claude Fable 583.0%
Published Jun 25, 2026Top 10 →ReasoningTIGER-LabMMLU-Pro12,000 harder, ten-option multiple-choice questions across 14 academic subjects.
  1. 1Gemini 3.1 Pro Preview91.2%
  2. 2Gemini 3 Pro90.1%
  3. 3o189.3%
Stale since Mar 11, 2026Top 10 →CodingLMArenaLMArena WebDev ArenaPeople compare two web apps built by anonymous models from the same prompt and vote for the better one.
  1. 1Claude Opus 5.51,814
  2. 2Claude Sonnet 5.51,772
  3. 3GPT-6 Astra1,770
Published Nov 6, 2026Top 10 →CodingScale AISWE-Bench ProLong-horizon software engineering tasks from real repositories: the public V2 set (relaunched September 2026), resolve rate on the Hard split the board opens on.
  1. 1Claude Opus 598.0%
  2. 2Claude Fable 5.192.2%
  3. 3Gemini 4 Argon92.2%
Published Nov 9, 2026Top 10 →CodingSWE-benchSWE-bench Verified500 human-verified GitHub issues. Bash-only runs with the same minimal agent, so models compare like for like.
  1. 1Claude Opus 4.576.8%
  2. 2Gemini 3 Flash75.8%
  3. 2MiniMax-M2.575.8%
Stale since Feb 26, 2026Top 10 →CodingAiderAider Polyglot225 hard Exercism exercises in C++, Go, Java, JavaScript, Python and Rust, edited through Aider.
  1. 1GPT-588.0%
  2. 2o3-pro84.9%
  3. 3Gemini 2.5 Pro (Jun 2025)83.1%
Stale since Oct 3, 2025Top 10 →Agents and toolsUC Berkeley GorillaBerkeley Function Calling LeaderboardFunction and tool calling: single, parallel, multi-turn, web search and memory, v4.
  1. 1Claude Opus 4.577.5%
  2. 2Claude Sonnet 4.573.2%
  3. 3Gemini 3 Pro72.5%
Stale since Apr 12, 2026Top 10 →Agents and toolsScale AIMCP AtlasReal tasks over 36 MCP servers and 220 tools, three to six tool calls each (the 500-task public subset).
  1. 1Gemini 4 Argon90.7%
  2. 2Muse Spark 1.188.1%
  3. 3Claude Fable 5.187.2%
Published Nov 9, 2026Top 10 →Agents and toolsLMArenaLMArena Search ArenaBlind votes on answers from search-grounded models.
  1. 1GPT-5.6 Sol1,255
  2. 2GPT-6.1 Sol1,247
  3. 3GPT-6 Astra1,245
Published Nov 6, 2026Top 10 →VisionLMArenaLMArena Vision ArenaBlind votes on conversations that include images. Style control on.
  1. 1Claude Fable 51,309
  2. 2Qwen3.8 Max (0902)1,300
  3. 3Claude Opus 4.71,299
Published Nov 6, 2026Top 10 →VisionLMArenaLMArena Document ArenaBlind votes on answers about uploaded documents and PDFs.
  1. 1Claude Fable 51,499
  2. 2Claude Fable 5.11,499
  3. 3Claude Opus 4.61,496
Published Nov 6, 2026Top 10 →ImageLMArenaLMArena Text-to-ImageBlind votes between two images generated from the same prompt.
  1. 1GPT Image 2.5 Sunburst1,429
  2. 2GPT Image 2.6 Sunburst1,409
  3. 3GPT Image 2.5 Flare1,402
Published Nov 6, 2026Top 10 →ImageLMArenaLMArena Image EditBlind votes between two edits of the same image and instruction.
  1. 1GPT Image 2.5 Sunburst1,517
  2. 2GPT Image 2.5 Flare1,483
  3. 3GPT Image 2.6 Sunburst1,476
Published Nov 6, 2026Top 10 →VideoLMArenaLMArena Text-to-VideoBlind votes between two videos generated from the same prompt.
  1. 1Gemini Omni 1.2 Flash1,541
  2. 2Gemini Omni 1.1 Flash1,514
  3. 3Gemini Omni Flash1,509
Published Nov 6, 2026Top 10 →VideoLMArenaLMArena Image-to-VideoBlind votes between two videos animated from the same image.
  1. 1Gemini Omni 1.2 Flash1,502
  2. 2MiniMax H31,493
  3. 3Dreamina Seedance 2.51,488
Published Nov 6, 2026Top 10 →SpeechArtificial AnalysisArtificial Analysis Speech ArenaBlind votes between two text-to-speech voices reading the same text.
  1. 1Eleven V41,339
  2. 2Eleven V4 Turbo1,335
  3. 3Qwen Audio 3.1 TTS Plus1,303
Published Nov 9, 2026Top 10 →SpeechHugging FaceOpen ASR LeaderboardSpeech recognition word error rate across eight public English test sets. Lower is better.
  1. 1scribe v2.1 pro (zoom)4.87 WER
  2. 2speech-v3.1 (OtterAI)5.21 WER
  3. 3azure-speech-09-2026 (microsoft)5.55 WER
Published Nov 6, 2026Top 10 →EmbeddingsMTEBMTEB Multilingual v2Text embeddings across 131 tasks and 250+ languages: retrieval, clustering, classification, similarity.
  1. 1harrier-oss-v1-27b (microsoft)74.3%
  2. 2KaLM-Embedding-Gemma3-12B-2511 (tencent)72.3%
  3. 3llama-embed-nemotron-8b (nvidia)69.5%
Published Nov 8, 2026Top 10 →SpeedArtificial AnalysisArtificial Analysis Output SpeedMedian output tokens per second on the first-party API, measured over the past two weeks.
  1. 1MiMo-V2.6-Flash64 tok/s
  2. 2Qwen3.8 Max (0902)60 tok/s
  3. 3Qwen3.8 27B44 tok/s
Published Nov 7, 2026Top 10 →PriceArtificial AnalysisArtificial Analysis PriceBlended price per million tokens (3 input : 1 output), among models scoring 30+ on the Intelligence Index.
  1. 1Claude Haiku 5.5$0.20
  2. 2Inkling 2 Small$0.47
  3. 3DeepSeek V4.1 Flash$0.53
Published Nov 9, 2026Top 10 →UsageOpenRouterOpenRouter weekly usageTokens sent to each model through OpenRouter in the latest week.
  1. 1DeepSeek V4.1 Flash56.0T tok
  2. 2GLM 5.3 Flash13.2T tok
  3. 3Hy4 preview12.8T tok
Published Nov 8, 2026Top 10 →ReasoningHugging FaceHugging Face Open LLM LeaderboardOpen-weights models on six academic benchmarks. Retired: results stopped updating in March 2025.Retired 2025-03-20Top 10 →