AI Agent Benchmark Results — All Leaderboards
Browse 370 sourced results across 14 AI agent benchmarks — WebVoyager, WebArena, OSWorld, SWE-bench Verified, GAIA, BrowseComp, and more. Every row links to the original paper, model card, repository, or launch post. Filter by category, benchmark, or search by agent or organization name.
Need methodology, evaluator details, and example tasks before comparing scores? Open the individual benchmark hub page for any benchmark below.
Scores on this page are not directly comparable across benchmarks.
This index puts every result in one place for browsing — it does not normalize methodology. Each benchmark uses its own task set, evaluator, scoring metric, and scope (model vs. agent). A 90% on one benchmark does not mean the same thing as a 90% on another. Compare scores within a single benchmark using the filters below, and read each benchmark's methodology notes on its dedicated page before drawing conclusions.
Category
Benchmark
Results
Sorted by reported date| Benchmark | Rank | System / Submission | Score | Organization | Reported | Source |
|---|---|---|---|---|---|---|
| WebArena | #16 | Nemotron 3.5 Bolt 74B-A9B-24GB
ⓘ
| 50.86% | NVIDIA | 2026-10-15 | Source |
| WebArena | #17 | Nemotron 3.5 Bolt 74B-A9B-BF16
ⓘ
| 50.25% | NVIDIA | 2026-10-15 | Source |
| BrowseComp | #38 | Nemotron 3.5 Bolt 74B-A9B-BF16
ⓘ
| 70.00% | NVIDIA | 2026-10-15 | Source |
| BrowseComp | #41 | Nemotron 3.5 Bolt 74B-A9B-24GB
ⓘ
| 66.75% | NVIDIA | 2026-10-15 | Source |
| DRACO | #6 | Claude Haiku 5.5
ⓘ
| 81.5% | Anthropic | 2026-10-07 | Source |
| Online-Mind2Web | #5 | CUAWright + GPT-5.5
ⓘ
| 89.4% | Microsoft Research / Academic Research | 2026-10-02 | Source |
| OSWorld 2.0 | #6 | Gemini 4 Argon
ⓘ
| 69.2% | 2026-09-30 | Source | |
| BrowseComp | #37 | Traverse-35B
ⓘ
| 72.83% | ByteDance / Peking University | 2026-09-30 | Source |
| OSWorld 2.0 | #5 | GPT-6.1 Sol
ⓘ
| 70.5% | OpenAI | 2026-09-29 | Source |
| BrowseComp | #17 | AREX-2
ⓘ
| 84.0% | Beijing Academy of Artificial Intelligence | 2026-09-29 | Source |
| DRACO | #3 | Claude Sonnet 5.5
ⓘ
| 87.0% | Anthropic | 2026-09-28 | Source |
| DRACO | #2 | Claude Opus 5.5
ⓘ
| 87.4% | Anthropic | 2026-09-22 | Source |
| OSWorld | #6 | MiMo-V2.6-Pro
ⓘ
| 82.0% | Xiaomi | 2026-09-22 | Source |
| OSWorld | #8 | MiMo-V2.6-Flash
ⓘ
| 80.8% | Xiaomi | 2026-09-22 | Source |
| BrowseComp | #2 | GPT-6 Astra
ⓘ
| 91.5% | OpenAI | 2026-09-03 | Source |
| OSWorld 2.0 | #3 | GPT-6 Astra
ⓘ
| 72.6% | OpenAI | 2026-09-03 | Source |
| OSWorld 2.0 | #10 | GPT-5.6 Sol (offline subset)
ⓘ
| 65.7% | OpenAI | 2026-09-03 | Source |
| OSWorld 2.0 | #8 | Muse Spark 1.3
ⓘ
| 66.9% | Meta | 2026-09-02 | Source |
| SWE-bench Verified | #1 | Claude Opus 5 (Vals.ai run)
ⓘ
| 97.00% | Anthropic | 2026-09-01 | Source |
| OSWorld 2.0 | #1 | Claude Fable 5.1
ⓘ
| 77.9% | Anthropic | 2026-09-01 | Source |
| τ-bench | #5 | Claude Fable 5.1 (OpenRouter run)
ⓘ
| 79.3% | Anthropic | 2026-09 | Source |
| τ-bench | #6 | Claude Opus 5 (OpenRouter run)
ⓘ
| 78.0% | Anthropic | 2026-09 | Source |
| OSWorld 2.0 | #7 | Claude Opus 5 (Snorkel run)
ⓘ
| 68.31% | Snorkel AI | 2026-09 | Source |
| OSWorld 2.0 | #11 | GPT-5.6 Sol (Snorkel run)
ⓘ
| 62.72% | Snorkel AI | 2026-09 | Source |
| OSWorld 2.0 | #2 | Simular Sai
ⓘ
| 73.0% | Simular AI | 2026-08-28 | Source |
| OSWorld 2.0 | #15 | Qwen3.8-Flash-Next
ⓘ
| 52.3% | Alibaba | 2026-08-26 | Source |
| Online-Mind2Web | #11 | Wuying-Browser-Agent-27B
ⓘ
| 66.7% | Alibaba Cloud | 2026-08-18 | Source |
| WebVoyager | #15 | Wuying-Browser-Agent-27B
ⓘ
| 80.6% | Alibaba Cloud | 2026-08-18 | Source |
| OSWorld | #1 | Qwen3.8-Max
ⓘ
| 86.1% | Alibaba | 2026-08-03 | Source |
| DRACO | #1 | Claude Opus 5
ⓘ
| 88.6% | Anthropic | 2026-07-24 | Source |
| SWE-bench Verified | #2 | Claude Opus 5
ⓘ
| 96.0% | Anthropic | 2026-07-24 | Source |
| BrowseComp | #4 | Claude Opus 5
ⓘ
| 90.8% | Anthropic | 2026-07-24 | Source |
| OSWorld 2.0 | #4 | Claude Opus 5
ⓘ
| 70.6% | Anthropic | 2026-07-24 | Source |
| OSWorld 2.0 | #9 | Claude Fable 5
ⓘ
| 66.1% | Anthropic | 2026-07-24 | Source |
| OSWorld 2.0 | #13 | Claude Opus 4.8 (Opus 5 card table)
ⓘ
| 55.7% | Anthropic | 2026-07-24 | Source |
| BrowseComp | #3 | Kimi K3
ⓘ
| 91.2% | Moonshot AI | 2026-07-16 | Source |
| BrowseComp | #1 | GPT-5.6 Sol Ultra
ⓘ
| 92.2% | OpenAI | 2026-07-09 | Source |
| BrowseComp | #5 | GPT-5.6 Sol
ⓘ
| 90.4% | OpenAI | 2026-07-09 | Source |
| OSWorld 2.0 | #12 | GPT-5.6 Sol
ⓘ
| 62.6% | OpenAI | 2026-07-09 | Source |
| SWE-bench Verified | #8 | GPT-5.6 Sol
ⓘ
| 82.2% | OpenAI | 2026-07 | Source |
| BrowseComp | #9 | Claude Fable 5
ⓘ
| 88.0% | Anthropic | 2026-07 | Source |
| SWE-bench Verified | #15 | GLM-5.2
ⓘ
| 80.0% | Zhipu AI | 2026-07 | Source |
| SWE-bench Verified | #23 | Inkling
ⓘ
| 77.6% | Thinking Machines | 2026-07 | Source |
| SWE-bench Verified | #24 | Kimi K2.5
ⓘ
| 76.8% | Moonshot AI | 2026-07 | Source |
| SWE-bench Verified | #25 | Nemotron 3 Ultra
ⓘ
| 70.7% | NVIDIA | 2026-07 | Source |
| BrowseComp | #28 | Inkling
ⓘ
| 77.1% | Thinking Machines | 2026-07 | Source |
| OSWorld | #7 | Claude Sonnet 5
ⓘ
| 81.2% | Anthropic | 2026-06-30 | Source |
| OSWorld | #9 | Claude Sonnet 4.6
ⓘ
| 78.5% | Anthropic | 2026-06-30 | Source |
| BrowseComp | #14 | Claude Sonnet 5
ⓘ
| 84.7% | Anthropic | 2026-06-30 | Source |
| BrowseComp | #30 | Claude Sonnet 4.6
ⓘ
| 76.2% | Anthropic | 2026-06-30 | Source |
| BrowseComp | #32 | Agents-A1
ⓘ
| 75.5% | Academic Research | 2026-06-29 | Source |
| BrowseComp | #33 | Agents-A1-SFT
ⓘ
| 74.6% | Academic Research | 2026-06-29 | Source |
| BrowseComp | #34 | Exa Agent High
ⓘ
| 74.0% | Exa | 2026-06-16 | Source |
| BrowseComp | #55 | Exa Agent Medium
ⓘ
| 52.8% | Exa | 2026-06-16 | Source |
| BrowseComp | #63 | Perplexity Agent Advanced
ⓘ
| 49.0% | Perplexity | 2026-06-16 | Source |
| BrowseComp | #66 | Perplexity Agent Deep
ⓘ
| 47.0% | Perplexity | 2026-06-16 | Source |
| BrowseComp | #73 | Exa Agent Low
ⓘ
| 43.5% | Exa | 2026-06-16 | Source |
| BrowseComp | #89 | Perplexity Agent Pro
ⓘ
| 20.0% | Perplexity | 2026-06-16 | Source |
| BrowseComp | #96 | Perplexity Agent Fast
ⓘ
| 10.0% | Perplexity | 2026-06-16 | Source |
| DRACO | #12 | Claude Fable 5
ⓘ
| 65.3% | Anthropic | 2026-06-12 | Source |
| WebArena | #11 | WebChallenger (GLM-4-32B)
ⓘ
| 56.3% | Independent Research | 2026-06-09 | Source |
| Online-Mind2Web | #19 | WebChallenger (GLM-4-32B)
ⓘ
| 51.0% | Independent Research | 2026-06-09 | Source |
| WebVoyager | #1 | browser-control (Fable 5)
ⓘ
| 99.19% | Om Labs | 2026-06 | Source |
| HealthAdminBench | #1 | Claude Mythos 5 (browser-use)
ⓘ
| 51.9% | Anthropic | 2026-06 | Source |
| HealthAdminBench | #1 | Claude Opus 4.8 (browser-use)
ⓘ
| 51.9% | Anthropic | 2026-06 | Source |
| OSWorld | #2 | Claude Mythos Preview
ⓘ
| 85.4% | Anthropic | 2026-06 | Source |
| SWE-bench Verified | #3 | Claude Mythos 5
ⓘ
| 95.5% | Anthropic | 2026-06 | Source |
| OSWorld | #3 | Claude Mythos 5
ⓘ
| 85.0% | Anthropic | 2026-06 | Source |
| HealthAdminBench | #3 | Claude Mythos Preview (browser-use)
ⓘ
| 47.4% | Anthropic | 2026-06 | Source |
| DRACO | #4 | Claude Mythos 5
ⓘ
| 86.4% | Anthropic | 2026-06 | Source |
| SWE-bench Verified | #4 | Claude Fable 5
ⓘ
| 95.0% | Anthropic | 2026-06 | Source |
| OSWorld | #4 | Claude Fable 5
ⓘ
| 85.0% | Anthropic | 2026-06 | Source |
| HealthAdminBench | #4 | Claude Sonnet 4.6 (browser-use)
ⓘ
| 45.2% | Anthropic | 2026-06 | Source |
| DRACO | #5 | Claude Mythos Preview
ⓘ
| 83.6% | Anthropic | 2026-06 | Source |
| WebVoyager | #6 | browser-control (Opus 4.8)
ⓘ
| 90.41% | browser-control | 2026-06 | Source |
| DRACO | #7 | Claude Opus 4.8
ⓘ
| 80.4% | Anthropic | 2026-06 | Source |
| BrowseComp | #9 | Claude Mythos 5
ⓘ
| 88.0% | Anthropic | 2026-06 | Source |
| OSWorld 2.0 | #14 | Claude Opus 4.8 (batched tools)
ⓘ
| 54.8% | Anthropic | 2026-06 | Source |
| OSWorld 2.0 | #16 | GPT-5.5 (batched tools)
ⓘ
| 49.5% | OpenAI | 2026-06 | Source |
| OSWorld 2.0 | #17 | Claude Opus 4.8
ⓘ
| 49.33% | Anthropic | 2026-06 | Source |
| OSWorld 2.0 | #18 | Claude Opus 4.7
ⓘ
| 49.1% | Anthropic | 2026-06 | Source |
| OSWorld 2.0 | #19 | Claude Opus 4.7 (batched tools)
ⓘ
| 48.91% | Anthropic | 2026-06 | Source |
| OSWorld 2.0 | #20 | Claude Sonnet 4.6 (max thinking)
ⓘ
| 41.5% | Anthropic | 2026-06 | Source |
| OSWorld 2.0 | #21 | Claude Sonnet 4.6 (medium thinking)
ⓘ
| 33.9% | Anthropic | 2026-06 | Source |
| OSWorld 2.0 | #22 | MiniMax M3
ⓘ
| 22.3% | MiniMax | 2026-06 | Source |
| OSWorld 2.0 | #23 | Kimi 2.6
ⓘ
| 22.1% | Moonshot AI | 2026-06 | Source |
| OSWorld 2.0 | #24 | Qwen 3.7-Plus
ⓘ
| 21.5% | Alibaba | 2026-06 | Source |
| DRACO | #9 | MiniMax M3
ⓘ
| 73.23% | MiniMax | 2026-05-31 | Source |
| OSWorld | #5 | Claude Opus 4.8
ⓘ
| 83.4% | Anthropic | 2026-05-28 | Source |
| SWE-bench Verified | #6 | Claude Opus 4.8
ⓘ
| 88.6% | Anthropic | 2026-05-28 | Source |
| BrowseComp | #16 | Claude Opus 4.8
ⓘ
| 84.3% | Anthropic | 2026-05-28 | Source |
| BrowseComp | #45 | LongSeeker
ⓘ
| 61.5% | Academic Research | 2026-05-06 | Source |
| BrowseComp | #67 | OpenSeeker-v2
ⓘ
| 46.0% | PolarSeeker | 2026-05-05 | Source |
| SWE-bench Verified | #11 | DeepSeek-V4-Pro-Max
ⓘ
| 80.6% | DeepSeek | 2026-04-24 | Source |
| SWE-bench Verified | #18 | DeepSeek-V4-Flash-Max
ⓘ
| 79.0% | DeepSeek | 2026-04-24 | Source |
| BrowseComp | #19 | DeepSeek-V4-Pro-Max
ⓘ
| 83.4% | DeepSeek | 2026-04-24 | Source |
| BrowseComp | #35 | DeepSeek-V4-Flash-Max
ⓘ
| 73.2% | DeepSeek | 2026-04-24 | Source |
| BrowseComp | #54 | Parallel Basic + GPT-5.4 harness
ⓘ
| 53.0% | Parallel | 2026-04-21 | Source |
| BrowseComp | #59 | Parallel Advanced + GPT-5.4 harness
ⓘ
| 51.0% | Parallel | 2026-04-21 | Source |
| BrowseComp | #65 | MiroThinker v1.0-72B
ⓘ
| 47.1% | MiroMind | 2026-04-21 | Source |
| BrowseComp | #76 | Tavily + GPT-5.4 harness
ⓘ
| 42.0% | Tavily | 2026-04-21 | Source |
| BrowseComp | #78 | Exa + GPT-5.4 harness
ⓘ
| 40.0% | Exa | 2026-04-21 | Source |
| BrowseComp | #12 | Kimi K2.6
ⓘ
| 86.3% | Moonshot AI | 2026-04-20 | Source |
| SWE-bench Verified | #13 | Kimi K2.6
ⓘ
| 80.2% | Moonshot AI | 2026-04-20 | Source |
| SWE-bench Verified | #7 | Claude Opus 4.7
ⓘ
| 87.6% | Anthropic | 2026-04-16 | Source |
| DRACO | #8 | Claude Opus 4.7
ⓘ
| 77.7% | Anthropic | 2026-04-16 | Source |
| BrowseComp | #21 | Claude Opus 4.7
ⓘ
| 79.3% | Anthropic | 2026-04-16 | Source |
| ClawBench | #1 | Claude Sonnet 4.6
ⓘ
| 33.3% | Anthropic | 2026-04-09 | Source |
| ClawBench | #2 | GLM-5
ⓘ
| 24.2% | Z.ai | 2026-04-09 | Source |
| ClawBench | #3 | Gemini 3 Flash
ⓘ
| 19.0% | 2026-04-09 | Source | |
| ClawBench | #4 | Claude Haiku 4.5
ⓘ
| 18.3% | Anthropic | 2026-04-09 | Source |
| SWE-bench Verified | #5 | Claude Mythos Preview
ⓘ
| 93.9% | Anthropic | 2026-04-09 | Source |
| ClawBench | #5 | GPT-5.4
ⓘ
| 6.5% | OpenAI | 2026-04-09 | Source |
| ClawBench | #6 | Gemini 3.1 Flash Lite
ⓘ
| 3.3% | 2026-04-09 | Source | |
| ClawBench | #7 | Kimi K2.5
ⓘ
| 0.7% | Moonshot AI | 2026-04-09 | Source |
| BrowseComp | #11 | Claude Mythos Preview
ⓘ
| 86.9% | Anthropic | 2026-04-07 | Source |
| BrowseComp | #18 | Claude Opus 4.6
ⓘ
| 83.7% | Anthropic | 2026-04-07 | Source |
| SWE-bench Verified | #19 | Qwen3.6 Plus
ⓘ
| 78.8% | Alibaba Cloud / Qwen Team | 2026-04-02 | Source |
| HealthAdminBench | #5 | Claude Opus 4.6 CUA
ⓘ
| 36.3% | Anthropic | 2026-04 | Source |
| BrowseComp | #6 | GPT-5.5 Pro
ⓘ
| 90.1% | OpenAI | 2026-04 | Source |
| HealthAdminBench | #6 | GPT-5.4 CUA
ⓘ
| 26.7% | OpenAI | 2026-04 | Source |
| BrowseComp | #7 | GPT-5.4 Pro
ⓘ
| 89.3% | OpenAI | 2026-04 | Source |
| HealthAdminBench | #7 | Kimi K2.5
ⓘ
| 15.6% | Moonshot AI | 2026-04 | Source |
| HealthAdminBench | #8 | Claude Opus 4.6
ⓘ
| 14.8% | Anthropic | 2026-04 | Source |
| WebVoyager | #9 | GLM-5V-Turbo
ⓘ
| 88.5% | Z.ai | 2026-04 | Source |
| HealthAdminBench | #9 | Qwen 3.5
ⓘ
| 13.3% | Alibaba | 2026-04 | Source |
| HealthAdminBench | #10 | Gemini 3.1 Pro
ⓘ
| 11.9% | 2026-04 | Source | |
| HealthAdminBench | #11 | GPT-5.4
ⓘ
| 5.9% | OpenAI | 2026-04 | Source |
| BrowseComp | #15 | GPT-5.5
ⓘ
| 84.4% | OpenAI | 2026-04 | Source |
| OSWorld | #15 | GLM-5V-Turbo
ⓘ
| 62.3% | Zhipu AI | 2026-04-01 | Source |
| WebArena | #19 | WebUncertainty + GPT-4-Turbo
ⓘ
| 46.9% | Academic Research | 2026-04 | Source |
| BrowseComp | #21 | GLM-5.1
ⓘ
| 79.3% | Zhipu AI | 2026-04 | Source |
| WebArena | #21 | A3-Qwen3.5-9B
ⓘ
| 42.1% | McGill NLP | 2026-04 | Source |
| BrowseComp | #29 | MiniMax M2.5
ⓘ
| 76.3% | MiniMax | 2026-04 | Source |
| Online-Mind2Web | #1 | Browser Use Cloud (bu-max)
ⓘ
| 97.0% | Browser-Use | 2026-03-25 | Source |
| Online-Mind2Web | #12 | Stagehand (Gemini 2.5 CU)
ⓘ
| 65.0% | Browserbase | 2026-03-25 | Source |
| Online-Mind2Web | #17 | Stagehand (Sonnet 4.5)
ⓘ
| 55.0% | Browserbase | 2026-03-25 | Source |
| SWE-bench Verified | #20 | MiMo-V2-Pro
ⓘ
| 78.0% | Xiaomi | 2026-03-18 | Source |
| BrowseComp | #8 | MiroThinker-H1
ⓘ
| 88.2% | MiroMind | 2026-03-16 | Source |
| BrowseComp | #85 | OpenSeeker
ⓘ
| 29.5% | PolarSeeker | 2026-03-16 | Source |
| GAIA | #1 | OPS-Agentic-Search
ⓘ
| 92.36% | Alibaba Cloud | 2026-03-11 | Source |
| GAIA | #1 | openJiuwen-deepagent
ⓘ
| 92.36% | Suzhou AI Lab / Shuqian Tech | 2026-03-11 | Source |
| BrowseComp | #82 | Nemotron 3 Super (120B A12B)
ⓘ
| 31.3% | NVIDIA | 2026-03-11 | Source |
| BrowseComp | #62 | Sarvam-105B
ⓘ
| 49.5% | Sarvam AI | 2026-03-06 | Source |
| BrowseComp | #80 | Sarvam-30B
ⓘ
| 35.5% | Sarvam AI | 2026-03-06 | Source |
| Online-Mind2Web | #2 | GPT-5.4 Native Computer Use
ⓘ
| 93.0% | OpenAI | 2026-03-05 | Source |
| Online-Mind2Web | #9 | ChatGPT Atlas Agent Mode
ⓘ
| 71.0% | OpenAI | 2026-03-05 | Source |
| OSWorld | #11 | GPT-5.4
ⓘ
| 75.0% | OpenAI | 2026-03-05 | Source |
| BrowseComp | #20 | GPT-5.4
ⓘ
| 82.7% | OpenAI | 2026-03-05 | Source |
| BrowseComp | #26 | GPT-5.3-Codex
ⓘ
| 77.3% | OpenAI | 2026-03-05 | Source |
| Online-Mind2Web | #3 | ABP + Claude Opus 4.6
ⓘ
| 90.53% | theredsix | 2026-03-03 | Source |
| WebVoyager | #2 | Alumnium
ⓘ
| 98.5% | Alumnium | 2026-03 | Source |
| BrowseComp | #64 | SMTL
ⓘ
| 48.6% | Academic Research | 2026-02-27 | Source |
| SWE-bench Verified | #11 | Gemini 3.1 Pro
ⓘ
| 80.6% | Google DeepMind | 2026-02-19 | Source |
| BrowseComp | #13 | Gemini 3.1 Pro
ⓘ
| 85.9% | 2026-02-19 | Source | |
| BrowseComp | #49 | Gemini 3 Pro
ⓘ
| 59.2% | 2026-02-19 | Source | |
| SWE-bench Verified | #17 | Claude Sonnet 4.6
ⓘ
| 79.6% | Anthropic | 2026-02-17 | Source |
| τ-bench | #10 | Qwen3.5-397B-A17B
ⓘ
| 68.4% | Alibaba | 2026-02-16 | Source |
| BrowseComp | #26 | Seed 2.0 Pro
ⓘ
| 77.3% | ByteDance | 2026-02-15 | Source |
| Online-Mind2Web | #4 | TinyFish
ⓘ
| 90.0% | TinyFish AI | 2026-02-12 | Source |
| DRACO | #10 | Perplexity Deep Research (Opus 4.6)
ⓘ
| 70.5% | Perplexity | 2026-02-12 | Source |
| DRACO | #11 | Perplexity Deep Research (Opus 4.5)
ⓘ
| 67.2% | Perplexity | 2026-02-12 | Source |
| DRACO | #13 | Claude Opus 4.6
ⓘ
| 59.8% | Anthropic | 2026-02-12 | Source |
| SWE-bench Verified | #13 | MiniMax M2.5
ⓘ
| 80.2% | MiniMax | 2026-02-12 | Source |
| DRACO | #14 | Gemini Deep Research
ⓘ
| 59.0% | 2026-02-12 | Source | |
| DRACO | #15 | OpenAI Deep Research (o3)
ⓘ
| 52.1% | OpenAI | 2026-02-12 | Source |
| DRACO | #16 | Claude Opus 4.5
ⓘ
| 46.7% | Anthropic | 2026-02-12 | Source |
| DRACO | #17 | OpenAI Deep Research (o4-mini)
ⓘ
| 41.9% | OpenAI | 2026-02-12 | Source |
| τ-bench | #13 | GLM-5
ⓘ
| 63.2% | Zhipu AI | 2026-02-11 | Source |
| SWE-bench Verified | #22 | GLM-5
ⓘ
| 77.8% | Zhipu AI | 2026-02-11 | Source |
| GAIA | #3 | openJiuwen-deepagent (GPT5/Gemini)
ⓘ
| 91.69% | openJiuwen | 2026-02-09 | Source |
| GAIA | #4 | Lemon Agent
ⓘ
| 91.36% | Lenovo CTO Org | 2026-02-06 | Source |
| SWE-bench Verified | #10 | Claude Opus 4.6
ⓘ
| 80.8% | Anthropic | 2026-02-05 | Source |
| OSWorld | #12 | Claude Opus 4.6
ⓘ
| 72.7% | Anthropic | 2026-02-05 | Source |
| WebArena | #1 | WebTactix (DeepSeek v3.2)
ⓘ
| 74.3% | WebTactix | 2026-02 | Source |
| WebArena | #8 | Kimi K2.5
ⓘ
| 58.9% | Moonshot AI | 2026-02 | Source |
| WebArena | #12 | Plan-MCTS + GPT-5-mini
ⓘ
| 55.3% | Academic Research | 2026-02 | Source |
| BrowseComp | #23 | Qwen3.5-397B-A17B
ⓘ
| 78.6% | Alibaba Cloud / Qwen Team | 2026-02 | Source |
| BrowseComp | #24 | Kimi K2.5
ⓘ
| 78.4% | Moonshot AI | 2026-02 | Source |
| BrowseComp | #31 | GLM-5
ⓘ
| 75.9% | Zhipu AI | 2026-02 | Source |
| BrowseComp | #39 | Step-3.5-Flash
ⓘ
| 69.0% | StepFun | 2026-02 | Source |
| BrowseComp | #43 | Qwen3.5-122B-A10B
ⓘ
| 63.8% | Alibaba Cloud / Qwen Team | 2026-02 | Source |
| BrowseComp | #46 | Qwen3.5-27B
ⓘ
| 61.0% | Alibaba Cloud / Qwen Team | 2026-02 | Source |
| BrowseComp | #46 | Qwen3.5-35B-A3B
ⓘ
| 61.0% | Alibaba Cloud / Qwen Team | 2026-02 | Source |
| BrowseComp | #75 | GLM-4.7-Flash
ⓘ
| 42.8% | Zhipu AI | 2026-02 | Source |
| τ-bench | #1 | Step-3.5-Flash
ⓘ
| 88.2% | StepFun | 2026-01-29 | Source |
| OSWorld | #14 | Kimi K2.5
ⓘ
| 63.3% | Moonshot AI | 2026-01-27 | Source |
| GAIA | #9 | ShawnAgent v3.1
ⓘ
| 89.37% | Independent | 2026-01-16 | Source |
| GAIA | #5 | JoinAI V2.2
ⓘ
| 90.7% | JoinAI-CMCC | 2026-01-14 | Source |
| GAIA | #11 | ShawnAgent v3.0
ⓘ
| 89.04% | Independent | 2026-01-14 | Source |
| GAIA | #7 | JoinAI V2.1
ⓘ
| 90.03% | JoinAI-CMCC | 2026-01-13 | Source |
| BrowseComp | #67 | WebAnchor-30B
ⓘ
| 46.0% | Academic Research | 2026-01-07 | Source |
| τ-bench | #3 | MiMo-V2-Flash
ⓘ
| 80.3% | Xiaomi | 2026-01-06 | Source |
| GAIA | #6 | Nemotron-ToolOrchestra
ⓘ
| 90.37% | NVIDIA | 2026-01-06 | Source |
| WebArena | #2 | OpAgent
ⓘ
| 71.6% | CodeFuse AI | 2026-01 | Source |
| BrowseComp | #36 | LongCat-Flash-Thinking-2601
ⓘ
| 73.1% | Meituan | 2026-01 | Source |
| BrowseComp | #50 | MiMo-V2-Flash
ⓘ
| 58.3% | Xiaomi | 2026-01 | Source |
| GAIA | #11 | JoinAI V2
ⓘ
| 89.04% | JoinAI-CMCC | 2025-12-28 | Source |
| GAIA | #7 | SU Zero (Shuqian Pro)
ⓘ
| 90.03% | Shuqian Tech | 2025-12-23 | Source |
| τ-bench | #2 | GLM-4.7
ⓘ
| 87.4% | Z.ai | 2025-12-22 | Source |
| τ-bench | #4 | GLM-4.7-Flash
ⓘ
| 79.5% | Z.ai | 2025-12-22 | Source |
| GAIA | #9 | HALO V1217-1
ⓘ
| 89.37% | Microsoft AI Asia | 2025-12-17 | Source |
| GAIA | #11 | HALO V1217
ⓘ
| 89.04% | Microsoft AI Asia | 2025-12-17 | Source |
| τ-bench | #11 | Gemini 3 Flash
ⓘ
| 67.8% | Google DeepMind | 2025-12-17 | Source |
| SWE-bench Verified | #20 | Gemini 3 Flash
ⓘ
| 78.0% | Google DeepMind | 2025-12-17 | Source |
| τ-bench | #9 | GPT-5.2
ⓘ
| 69.9% | OpenAI | 2025-12-11 | Source |
| GAIA | #11 | Su Zero + SQ Pro
ⓘ
| 89.04% | Suzhou AI Lab / Shuqian Tech | 2025-12-11 | Source |
| SWE-bench Verified | #15 | GPT-5.2
ⓘ
| 80.0% | OpenAI | 2025-12-11 | Source |
| BrowseComp | #25 | GPT-5.2 Pro
ⓘ
| 77.9% | OpenAI | 2025-12-11 | Source |
| BrowseComp | #42 | GPT-5.2
ⓘ
| 65.8% | OpenAI | 2025-12-11 | Source |
| GAIA | #16 | Su Zero + Shuqian Lite
ⓘ
| 87.38% | Suzhou AI Lab / Shuqian Tech | 2025-12-07 | Source |
| GAIA | #15 | Lemon Agent v1.0.8
ⓘ
| 88.37% | Lenovo CTO Org | 2025-12-04 | Source |
| WebArena | #3 | ColorBrowserAgent
ⓘ
| 71.2% | MadeAgents | 2025-12 | Source |
| Online-Mind2Web | #7 | OpenAGI Lux
ⓘ
| 83.6% | OpenAGI Foundation | 2025-12-01 | Source |
| WebArena | #13 | WebOperator + GPT-4o
ⓘ
| 54.6% | KAIST KAG NLP | 2025-12 | Source |
| BrowseComp | #40 | GLM-4.7
ⓘ
| 67.5% | Zhipu AI | 2025-12 | Source |
| BrowseComp | #44 | MiniMax M2.1
ⓘ
| 62.0% | MiniMax | 2025-12 | Source |
| BrowseComp | #57 | DeepSeek-V3.2
ⓘ
| 51.4% | DeepSeek | 2025-12-01 | Source |
| BrowseComp | #57 | DeepSeek-V3.2 (Thinking)
ⓘ
| 51.4% | DeepSeek | 2025-12-01 | Source |
| OSWorld | #21 | Qwen3-VL-235B-A22B-Instruct
ⓘ
| 31.6% | Alibaba | 2025-11-26 | Source |
| τ-bench | #8 | Claude Opus 4.5
ⓘ
| 70.2% | Anthropic | 2025-11-24 | Source |
| SWE-bench Verified | #9 | Claude Opus 4.5
ⓘ
| 80.9% | Anthropic | 2025-11-24 | Source |
| OSWorld | #13 | Claude Opus 4.5
ⓘ
| 66.3% | Anthropic | 2025-11-24 | Source |
| Online-Mind2Web | #8 | Navigator
ⓘ
| 78.7% | Yutori | 2025-11-19 | Source |
| Online-Mind2Web | #14 | Claude 4.0
ⓘ
| 61.0% | Anthropic | 2025-11-19 | Source |
| Online-Mind2Web | #17 | Claude 4.5
ⓘ
| 55.0% | Anthropic | 2025-11-19 | Source |
| τ-bench | #12 | Gemini 3 Pro
ⓘ
| 65.8% | Google DeepMind | 2025-11-18 | Source |
| BrowseComp | #48 | Kimi K2-Thinking-0905
ⓘ
| 60.2% | Moonshot AI | 2025-11 | Source |
| τ-bench | #7 | MiniMax M2
ⓘ
| 77.2% | MiniMax | 2025-10-27 | Source |
| WebArena | #4 | Claude Code + GBOX MCP
ⓘ
| 68.0% | GBOX AI | 2025-10-25 | Source |
| OSWorld | #10 | OSAgent
ⓘ
| 76.26% | TheAGI Company | 2025-10-23 | Source |
| OSWorld | #18 | Claude Haiku 4.5
ⓘ
| 50.7% | Anthropic | 2025-10-15 | Source |
| GAIA | #17 | Co-Sight v2.1.0
ⓘ
| 87.04% | ZTE-AICloud | 2025-10-13 | Source |
| GAIA | #18 | JoinAI V1.1
ⓘ
| 86.71% | JoinAI | 2025-10-09 | Source |
| BrowseComp | #81 | DeepMiner-32B
ⓘ
| 33.5% | Academic Research | 2025-10-09 | Source |
| Online-Mind2Web | #10 | Gemini 2.5 Computer Use
ⓘ
| 69.0% | Google DeepMind | 2025-10-07 | Source |
| AgentBench | #1 | AgentRL w/ Qwen2.5-32B-Instruct
ⓘ
| 70.4% | Tsinghua University | 2025-10-05 | Source |
| AgentBench | #2 | AgentRL w/ Qwen2.5-14B-Instruct
ⓘ
| 67.7% | Tsinghua University | 2025-10-05 | Source |
| AgentBench | #3 | AgentRL w/ GLM-4-9B-0414
ⓘ
| 65.0% | Tsinghua University | 2025-10-05 | Source |
| AgentBench | #4 | AgentRL w/ Qwen2.5-7B-Instruct
ⓘ
| 62.0% | Tsinghua University | 2025-10-05 | Source |
| AgentBench | #5 | AgentRL w/ Qwen2.5-3B-Instruct
ⓘ
| 60.0% | Tsinghua University | 2025-10-05 | Source |
| Aider | #12 | DeepSeek-V3.2-Exp (Reasoner)
ⓘ
| 74.2% | DeepSeek | 2025-10-03 | Source |
| WebVoyager | #3 | Surfer 2
ⓘ
| 97.1% | H Company | 2025-10 | Source |
| WebArena | #6 | Narada AI
ⓘ
| 64.2% | Narada AI | 2025-10 | Source |
| BrowseComp | #69 | GLM-4.6
ⓘ
| 45.1% | Zhipu AI | 2025-10 | Source |
| BrowseComp | #72 | MiniMax M2
ⓘ
| 44.0% | MiniMax | 2025-10 | Source |
| BrowseComp | #77 | DeepSeek-V3.2-Exp
ⓘ
| 40.1% | DeepSeek | 2025-09-30 | Source |
| AgentBench | #6 | Claude Sonnet 4.5
ⓘ
| 58.9% | Anthropic | 2025-09-29 | Source |
| AgentBench | #7 | Claude Sonnet 4.5 Thinking
ⓘ
| 58.3% | Anthropic | 2025-09-29 | Source |
| τ-bench | #14 | Claude Sonnet 4.5
ⓘ
| 62.9% | Anthropic | 2025-09-29 | Source |
| OSWorld | #16 | Claude Sonnet 4.5
ⓘ
| 61.4% | Anthropic | 2025-09-29 | Source |
| BrowseComp | #91 | InfoAgent
ⓘ
| 15.3% | Academic Research | 2025-09-29 | Source |
| BrowseComp | #71 | Grok 4 Fast
ⓘ
| 44.9% | xAI | 2025-09-22 | Source |
| BrowseComp | #74 | Tongyi DeepResearch
ⓘ
| 43.4% | Alibaba Cloud / Tongyi Lab | 2025-09-16 | Source |
| BrowseComp | #79 | AgentFounder-30B
ⓘ
| 39.9% | Alibaba Cloud / Tongyi Lab | 2025-09-16 | Source |
| OSWorld | #17 | UiPath Screen Agent w/ GPT-5
ⓘ
| 53.6% | UiPath | 2025-09-14 | Source |
| BrowseComp | #91 | DeepDive-32B
ⓘ
| 15.3% | THUDM / Tsinghua University | 2025-09-12 | Source |
| BrowseComp | #90 | WebExplorer-8B (RL)
ⓘ
| 15.7% | HKUST NLP Group | 2025-09-08 | Source |
| BrowseComp | #95 | WebSailor-32B
ⓘ
| 10.5% | Alibaba Cloud / Tongyi Lab | 2025-09-08 | Source |
| BrowseComp | #100 | WebSailor-7B
ⓘ
| 6.7% | Alibaba Cloud / Tongyi Lab | 2025-09-08 | Source |
| Online-Mind2Web | #6 | UI-TARS-2
ⓘ
| 88.2% | ByteDance / VLM-Research | 2025-09-02 | Source |
| WebArena | #5 | DeepSky Agent
ⓘ
| 66.9% | DeepSky | 2025-09 | Source |
| Aider | #2 | gpt-5 (medium)
ⓘ
| 86.7% | OpenAI | 2025-08-25 | Source |
| Aider | #5 | gpt-5 (low)
ⓘ
| 81.3% | OpenAI | 2025-08-25 | Source |
| Aider | #1 | gpt-5 (high)
ⓘ
| 88.0% | OpenAI | 2025-08-23 | Source |
| Online-Mind2Web | #15 | ACT-1-20250814
ⓘ
| 57.3% | Enhans | 2025-08-14 | Source |
| BrowseComp | #83 | BrowseMaster
ⓘ
| 30.0% | Academic Research | 2025-08-12 | Source |
| WebVoyager | #17 | WebSight
ⓘ
| 68% | Academic Research | 2025-08 | Source |
| BrowseComp | #51 | Parallel Ultra8x
ⓘ
| 58.0% | Parallel | 2025-08 | Source |
| BrowseComp | #52 | Parallel Ultra4x
ⓘ
| 56.0% | Parallel | 2025-08 | Source |
| BrowseComp | #53 | GPT-5
ⓘ
| 54.9% | OpenAI | 2025-08 | Source |
| BrowseComp | #59 | Parallel Ultra2x
ⓘ
| 51.0% | Parallel | 2025-08 | Source |
| BrowseComp | #70 | Parallel Ultra
ⓘ
| 45.0% | Parallel | 2025-08 | Source |
| BrowseComp | #83 | DeepSeek-V3.1
ⓘ
| 30.0% | DeepSeek | 2025-08 | Source |
| BrowseComp | #93 | Exa Research Pro
ⓘ
| 14.0% | Exa | 2025-08 | Source |
| BrowseComp | #99 | Claude Opus 4.1 (Parallel Task API benchmark)
ⓘ
| 7.0% | Anthropic | 2025-08 | Source |
| BrowseComp | #101 | Perplexity Sonar Deep Research
ⓘ
| 6.0% | Perplexity | 2025-08 | Source |
| Aider | #7 | grok-4 (high)
ⓘ
| 79.6% | xAI | 2025-07-11 | Source |
| Online-Mind2Web | #20 | ACT-1-20250703
ⓘ
| 45.7% | Enhans | 2025-07-03 | Source |
| WebVoyager | #4 | Magnitude
ⓘ
| 93.9% | Magnitude | 2025-07 | Source |
| BrowseComp | #87 | GLM-4.5
ⓘ
| 26.4% | Zhipu AI | 2025-07 | Source |
| BrowseComp | #88 | GLM-4.5-Air
ⓘ
| 21.3% | Zhipu AI | 2025-07 | Source |
| BrowseComp | #94 | WebSailor-72B
ⓘ
| 12.0% | Alibaba Cloud / Tongyi Lab | 2025-07 | Source |
| Aider | #3 | o3-pro (high)
ⓘ
| 84.9% | OpenAI | 2025-06-28 | Source |
| Aider | #9 | o3 (high) + gpt-4.1
ⓘ
| 78.2% | OpenAI | 2025-06-27 | Source |
| Aider | #5 | o3 (high)
ⓘ
| 81.3% | OpenAI | 2025-06-25 | Source |
| Aider | #10 | o3
ⓘ
| 76.9% | OpenAI | 2025-06-25 | Source |
| Aider | #4 | gemini-2.5-pro-preview-06-05 (32k think)
ⓘ
| 83.1% | 2025-06-06 | Source | |
| Aider | #8 | gemini-2.5-pro-preview-06-05 (default think)
ⓘ
| 79.1% | 2025-06-06 | Source | |
| WebVoyager | #5 | Surfer-H + Holo1
ⓘ
| 92.2% | H Company | 2025-06 | Source |
| WebArena | #22 | GUI-API Hybrid Agent
ⓘ
| 38.9% | Academic Research | 2025-06 | Source |
| WebArena | #26 | TTI
ⓘ
| 26.1% | Academic Research | 2025-06 | Source |
| BrowseComp | #98 | DeepSeek-R1-0528
ⓘ
| 8.9% | DeepSeek | 2025-05-28 | Source |
| Aider | #14 | claude-opus-4-20250514 (32k thinking)
ⓘ
| 72.0% | Anthropic | 2025-05-25 | Source |
| AgentBench | #8 | Claude Sonnet 4 Thinking
ⓘ
| 58.2% | Anthropic | 2025-05-22 | Source |
| AgentBench | #9 | Claude Sonnet 4
ⓘ
| 57.4% | Anthropic | 2025-05-22 | Source |
| Aider | #10 | Gemini 2.5 Pro Preview 05-06
ⓘ
| 76.9% | 2025-05-07 | Source | |
| Aider | #14 | o4-mini (high)
ⓘ
| 72.0% | OpenAI | 2025-04-16 | Source |
| BrowseComp | #61 | o3
ⓘ
| 49.7% | OpenAI | 2025-04-16 | Source |
| BrowseComp | #86 | o4-mini
ⓘ
| 28.3% | OpenAI | 2025-04-16 | Source |
| Aider | #13 | Gemini 2.5 Pro Preview 03-25
ⓘ
| 72.9% | 2025-04-12 | Source | |
| BrowseComp | #56 | OpenAI Deep Research
ⓘ
| 51.5% | OpenAI | 2025-04-10 | Source |
| BrowseComp | #97 | OpenAI o1
ⓘ
| 9.9% | OpenAI | 2025-04-10 | Source |
| BrowseComp | #102 | GPT-4o + browsing
ⓘ
| 1.9% | OpenAI | 2025-04-10 | Source |
| BrowseComp | #103 | GPT-4.5
ⓘ
| 0.9% | OpenAI | 2025-04-10 | Source |
| BrowseComp | #104 | GPT-4o
ⓘ
| 0.6% | OpenAI | 2025-04-10 | Source |
| Online-Mind2Web | #13 | OpenAI Operator
ⓘ
| 61.3% | OpenAI | 2025-04-02 | Source |
| WebVoyager | #7 | Browserable
ⓘ
| 90.4% | Browserable | 2025-04 | Source |
| WebVoyager | #12 | Notte
ⓘ
| 86.2% | Notte | 2025-04 | Source |
| Online-Mind2Web | #22 | HAL Leaderboard baseline (best open)
ⓘ
| ~30% | Princeton / OSU | 2025-04 | Source |
| OSWorld | #23 | Qwen2.5 VL 32B Instruct
ⓘ
| 5.9% | Alibaba Cloud / Qwen Team | 2025-03-24 | Source |
| OSWorld | #19 | Agent S2 + Claude 3.7
ⓘ
| 34.5% | Simular AI | 2025-03-12 | Source |
| GAIA | #19 | Manus
ⓘ
| 86.5% | Monica AI | 2025-03-06 | Source |
| WebVoyager | #18 | Runner H 0.1
ⓘ
| 67% | H Company | 2025-03 | Source |
| AgentBench | #10 | Claude Sonnet 3.7
ⓘ
| 53.2% | Anthropic | 2025-02-24 | Source |
| Online-Mind2Web | #16 | Claude Computer Use 3.7 (w/o thinking)
ⓘ
| 56.3% | Anthropic | 2025-02-24 | Source |
| WebArena | #7 | IBM CUGA
ⓘ
| 61.7% | IBM | 2025-02-17 | Source |
| GAIA | #20 | Deep Research (o3, cons@64)
ⓘ
| 72.57% | OpenAI | 2025-02-02 | Source |
| GAIA | #21 | Deep Research (o3)
ⓘ
| 67.36% | OpenAI | 2025-02-02 | Source |
| OSWorld | #22 | Qwen2.5 VL 72B Instruct
ⓘ
| 8.8% | Alibaba Cloud / Qwen Team | 2025-01-26 | Source |
| OSWorld | #20 | OpenAI Operator (CUA)
ⓘ
| 32.6% | OpenAI | 2025-01-23 | Source |
| WebArena | #9 | OpenAI Operator
ⓘ
| 58.1% | OpenAI | 2025-01 | Source |
| WebVoyager | #10 | Operator
ⓘ
| 87% | OpenAI | 2025-01 | Source |
| WebVoyager | #13 | Skyvern 2.0
ⓘ
| 85.85% | Skyvern | 2025-01 | Source |
| WebArena | #15 | AgentSymbiotic
ⓘ
| 52.1% | Academic Research | 2025-01 | Source |
| WebArena | #18 | Learn-by-Interact
ⓘ
| 48.0% | Academic Research | 2025-01 | Source |
| WebArena | #28 | AgentTrek-1.0-32B
ⓘ
| 22.4% | xLang AI | 2025-01 | Source |
| WebArena | #32 | NNetNav
ⓘ
| 16.3% | Stanford NLP | 2025-01 | Source |
| WebArena | #14 | ScribeAgent + GPT-4o
ⓘ
| 53.0% | Academic Research | 2024-12-24 | Source |
| WebVoyager | #8 | Browser Use
ⓘ
| 89.1% | Browser Use | 2024-12 | Source |
| WebVoyager | #14 | Project Mariner
ⓘ
| 83.5% | 2024-12 | Source | |
| Online-Mind2Web | #22 | Browser Use (gpt-4o)
ⓘ
| 30.0% | Browser Use | 2024-11-06 | Source |
| WebVoyager | #10 | Agent Kura
ⓘ
| 87.0% | Kura | 2024-11 | Source |
| WebVoyager | #20 | Anthropic Computer Use 3.5
ⓘ
| 56.0% | Anthropic | 2024-11 | Source |
| Online-Mind2Web | #24 | Claude Computer Use 3.5
ⓘ
| 29.0% | Anthropic | 2024-10-22 | Source |
| WebArena | #20 | AgentOccam-Judge
ⓘ
| 45.7% | Amazon Science | 2024-10 | Source |
| WebArena | #41 | Synatra-CodeLLama7b
ⓘ
| 6.28% | Academic Research | 2024-10 | Source |
| WebArena | #24 | Agent Workflow Memory
ⓘ
| 35.5% | Academic Research | 2024-09 | Source |
| WebArena | #10 | Jace.AI (AWA-1.5)
ⓘ
| 57.1% | Jace AI | 2024-08 | Source |
| WebArena | #23 | WebPilot
ⓘ
| 37.2% | Academic Research | 2024-08 | Source |
| Online-Mind2Web | #25 | Agent-E (gpt-4o)
ⓘ
| 28.0% | Emergence AI | 2024-07-16 | Source |
| WebVoyager | #16 | Agent-E
ⓘ
| 73.2% | Emergence AI | 2024-07 | Source |
| WebArena | #30 | GPT-4o + Tree Search
ⓘ
| 19.2% | Academic Research | 2024-06 | Source |
| WebArena | #34 | gpt-4o-2024-05-13
ⓘ
| 13.05% | OpenAI | 2024-05 | Source |
| WebArena | #36 | Patel et al. (Qwen1.5-72B, Mixture B)
ⓘ
| 9.36% | Academic Research | 2024-05 | Source |
| WebArena | #40 | Llama3-chat-70b
ⓘ
| 7.02% | Meta | 2024-04-02 | Source |
| WebVoyager | #21 | WILBUR
ⓘ
| 53% | Bardeen / UC Berkeley | 2024-04 | Source |
| WebArena | #25 | SteP
ⓘ
| 33.5% | ASAPP Research | 2024-04 | Source |
| WebArena | #27 | BrowserGym + GPT-4o
ⓘ
| 23.5% | ServiceNow Research | 2024-04 | Source |
| WebArena | #29 | GPT-4 + Auto Eval
ⓘ
| 20.2% | Academic Research | 2024-04 | Source |
| WebArena | #31 | AutoWebGLM
ⓘ
| 18.2% | THUDM | 2024-04 | Source |
| WebArena | #46 | Llama3-chat-8b
ⓘ
| 3.32% | Meta | 2024-04 | Source |
| WebArena | #43 | Agent-FLAN
ⓘ
| 4.68% | InternLM | 2024-03 | Source |
| WebArena | #47 | CodeAct Agent
ⓘ
| 2.3% | Academic Research | 2024-02 | Source |
| Online-Mind2Web | #21 | SeeAct (gpt-4o)
ⓘ
| 30.7% | OSU NLP | 2024-01-16 | Source |
| WebVoyager | #19 | WebVoyager
ⓘ
| 59.1% | Academic Research | 2024-01 | Source |
| WebVoyager | #22 | GPT-4 (All Tools)
ⓘ
| 30.8% | OpenAI | 2024-01 | Source |
| WebArena | #49 | Mixtral-8x7B
ⓘ
| 1.39% | Mistral AI | 2024-01 | Source |
| WebArena | #39 | Gemini Pro
ⓘ
| 7.12% | 2023-12 | Source | |
| WebArena | #42 | Lemur-chat-70b
ⓘ
| 5.3% | OpenLemur | 2023-10 | Source |
| WebArena | #45 | AgentLM-70b
ⓘ
| 3.81% | THUDM | 2023-10 | Source |
| WebArena | #48 | AgentLM-13b
ⓘ
| 1.6% | THUDM | 2023-10 | Source |
| WebArena | #50 | AgentLM-7b
ⓘ
| 0.74% | THUDM | 2023-10 | Source |
| WebArena | #51 | FireAct
ⓘ
| 0.25% | Academic Research | 2023-10 | Source |
| WebArena | #38 | Qwen-1.5-chat-72b
ⓘ
| 7.14% | Qwen | 2023-09 | Source |
| WebArena | #44 | CodeLlama-instruct-34b
ⓘ
| 4.06% | Meta | 2023-08 | Source |
| WebArena | #52 | CodeLlama-instruct-7b
ⓘ
| 0.0% | Meta | 2023-08 | Source |
| WebArena | #33 | gpt-4-0613 (no not-achievable hint)
ⓘ
| 14.9% | OpenAI | 2023-06 | Source |
| WebArena | #35 | gpt-4-0613 (with not-achievable hint)
ⓘ
| 11.7% | OpenAI | 2023-06 | Source |
| WebArena | #37 | gpt-3.5-turbo-16k-0613
ⓘ
| 8.87% | OpenAI | 2023-03 | Source |
No results match these filters. Clear filters or try a different search.
Scores come from public papers, model cards, repositories, and launch posts. Comparisons are most useful within a single benchmark — across benchmarks, evaluators, task sets, judges, attempt budgets, and tool access can differ, so treat cross-benchmark rankings as directional. Use the individual benchmark pages for methodology and interpretation notes.
Frequently asked questions
Are scores on this page comparable across benchmarks? + -
No. Each benchmark uses its own task set, evaluator, scoring metric, and scope (model vs. agent). A 90% on one benchmark and a 90% on another do not measure the same thing. Compare scores within a single benchmark, and read methodology notes on the individual benchmark pages before drawing conclusions.
Why is the default order by reported date rather than score? + -
Scores across different benchmarks aren't on the same scale, so a global score sort would be misleading. Sorting by reported date surfaces the freshest results first. When you filter to a single benchmark the order switches to within-benchmark rank, which is the meaningful comparison.
Are these scores independently verified? + -
Not always. Some rows are independently benchmarked and some are self-reported by the agent or model team. Every row links to its source (paper, model card, repository, or launch post). Use those links and any per-row notes to assess evidence level before drawing strong conclusions.
How is this index sourced and updated? + -
Rows are taken from public papers, model cards, repositories, and launch posts. New results appear as they are published — typically weekly. If you spot a missing agent or a stale score, open a pull request or issue on GitHub.
How do I add my agent or model to this list? + -
Open a pull request on GitHub. You need a publicly verifiable benchmark score, a link to the source (paper, post, or repo), and a homepage or repository for your agent.