# AI Browser Agent Leaderboards
Open reference for evaluating AI browser agents, computer-use systems, and coding agents across the public benchmarks teams actually compare on.
**Live site: [leaderboard.steel.dev](https://leaderboard.steel.dev)** — full rankings, methodology notes, and per-result detail.
Maintained by [Steel](https://steel.dev), the open-source browser API for AI agents.
## Top results by benchmark
The tables below show the current top entries on each tracked benchmark. Each section links to the full leaderboard with sources, methodology, and additional context.
<!-- LEADERBOARDS:START -->
### [WebVoyager](https://leaderboard.steel.dev/leaderboards/webvoyager)
Browser agents · Agent scope · 22 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | browser-control (Fable 5) **(new)** | Om Labs | [99.19%](https://github.com/omxyz/webvoyager) |
| 2 | Alumnium **(new)** | Alumnium | [98.5%](https://alumnium.ai/blog/webvoyager-benchmark/) |
| 3 | Surfer 2 | H Company | [97.1%](https://hcompany.ai/surfer-2) |
| 4 | Magnitude | Magnitude | [93.9%](https://github.com/magnitudedev/webvoyager) |
| 5 | Surfer-H + Holo1 | H Company | [92.2%](https://arxiv.org/pdf/2506.02865) |
[See all 22 entries →](https://leaderboard.steel.dev/leaderboards/webvoyager)
---
### [BrowseComp](https://leaderboard.steel.dev/leaderboards/browsecomp)
Research/search · Mixed scope · 104 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | GPT-5.6 Sol Ultra **(new)** | OpenAI | [92.2%](https://openai.com/index/gpt-5-6/) |
| 2 | GPT-6 Astra **(new)** | OpenAI | [91.5%](https://openai.com/index/gpt-6-astra/) |
| 3 | Kimi K3 **(new)** | Moonshot AI | [91.2%](https://www.kimi.com/blog/kimi-k3) |
| 4 | Claude Opus 5 **(new)** | Anthropic | [90.8%](https://www-cdn.anthropic.com/ceaf5c7ff2783855203fde8208ec311252dced5b/Claude%20Opus%205%20System%20Card.pdf) |
| 5 | GPT-5.6 Sol **(new)** | OpenAI | [90.4%](https://openai.com/index/gpt-5-6/) |
[See all 104 entries →](https://leaderboard.steel.dev/leaderboards/browsecomp)
---
### [DRACO](https://leaderboard.steel.dev/leaderboards/draco)
Research/search · Mixed scope · 17 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | Claude Opus 5 **(new)** | Anthropic | [88.6%](https://www-cdn.anthropic.com/ceaf5c7ff2783855203fde8208ec311252dced5b/Claude%20Opus%205%20System%20Card.pdf) |
| 2 | Claude Opus 5.5 **(new)** | Anthropic | [87.4%](https://www.anthropic.com/claude-opus-5-5-system-card) |
| 3 | Claude Sonnet 5.5 **(new)** | Anthropic | [87.0%](https://www.anthropic.com/claude-sonnet-5-5-system-card) |
| 4 | Claude Mythos 5 **(new)** | Anthropic | [86.4%](https://www.anthropic.com/claude-fable-5-mythos-5-system-card) |
| 5 | Claude Mythos Preview | Anthropic | [83.6%](https://www.anthropic.com/claude-fable-5-mythos-5-system-card) |
[See all 17 entries →](https://leaderboard.steel.dev/leaderboards/draco)
---
### [WebArena](https://leaderboard.steel.dev/leaderboards/webarena)
Browser agents · Agent scope · 52 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | WebTactix (DeepSeek v3.2) | WebTactix | [74.3%](https://paper-submission-anoymous.github.io/webtactix_introduction/) |
| 2 | OpAgent | CodeFuse AI | [71.6%](https://github.com/codefuse-ai/OpAgent) |
| 3 | ColorBrowserAgent | MadeAgents | [71.2%](https://arxiv.org/abs/2601.07262) |
| 4 | Claude Code + GBOX MCP | GBOX AI | [68.0%](https://github.com/babelcloud/webarena/blob/main/gbox/report.md) |
| 5 | DeepSky Agent | DeepSky | [66.9%](https://docs.google.com/spreadsheets/d/1M801lEpBbKSNwP-vDBkC_pF7LdyGU1f_ufZb_NWNBZQ) |
[See all 52 entries →](https://leaderboard.steel.dev/leaderboards/webarena)
---
### [SWE-bench Verified](https://leaderboard.steel.dev/leaderboards/swe-bench-verified)
Coding · Model scope · 25 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | Claude Opus 5 (Vals.ai run) **(new)** | Anthropic | [97.00%](https://vals.ai/benchmarks/swebench) |
| 2 | Claude Opus 5 **(new)** | Anthropic | [96.0%](https://www-cdn.anthropic.com/ceaf5c7ff2783855203fde8208ec311252dced5b/Claude%20Opus%205%20System%20Card.pdf) |
| 3 | Claude Mythos 5 **(new)** | Anthropic | [95.5%](https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf) |
| 4 | Claude Fable 5 **(new)** | Anthropic | [95.0%](https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf) |
| 5 | Claude Mythos Preview | Anthropic | [93.9%](https://www.mindstudio.ai/blog/claude-mythos-benchmark-results-swe-bench-agentic-coding) |
[See all 25 entries →](https://leaderboard.steel.dev/leaderboards/swe-bench-verified)
---
### [Aider](https://leaderboard.steel.dev/leaderboards/aider)
Coding · Model scope · 15 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | gpt-5 (high) | OpenAI | [88.0%](https://aider.chat/docs/leaderboards/) |
| 2 | gpt-5 (medium) | OpenAI | [86.7%](https://aider.chat/docs/leaderboards/) |
| 3 | o3-pro (high) | OpenAI | [84.9%](https://aider.chat/docs/leaderboards/) |
| 4 | gemini-2.5-pro-preview-06-05 (32k think) | Google | [83.1%](https://aider.chat/docs/leaderboards/) |
| 5 | o3 (high) | OpenAI | [81.3%](https://aider.chat/docs/leaderboards/) |
[See all 15 entries →](https://leaderboard.steel.dev/leaderboards/aider)
---
### [OSWorld](https://leaderboard.steel.dev/leaderboards/osworld)
Computer use · Agent scope · 23 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | Qwen3.8-Max **(new)** | Alibaba | [86.1%](https://qwen.ai/blog?id=qwen3.8) |
| 2 | Claude Mythos Preview **(new)** | Anthropic | [85.4%](https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf) |
| 3 | Claude Mythos 5 **(new)** | Anthropic | [85.0%](https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf) |
| 4 | Claude Fable 5 **(new)** | Anthropic | [85.0%](https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf) |
| 5 | Claude Opus 4.8 **(new)** | Anthropic | [83.4%](https://www.anthropic.com/news/claude-opus-4-8) |
[See all 23 entries →](https://leaderboard.steel.dev/leaderboards/osworld)
---
### [OSWorld 2.0](https://leaderboard.steel.dev/leaderboards/osworld-2)
Computer use · Agent scope · 24 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | Claude Fable 5.1 **(new)** | Anthropic | [77.9%](https://www-cdn.anthropic.com/0339e6a7c5c7b87f5c07798616dc32c215d14235/Claude%20Fable%205.1%20&%20Claude%20Mythos%205.1%20System%20Card.pdf) |
| 2 | Simular Sai **(new)** | Simular AI | [73.0%](https://www.simular.ai/articles/sai-tops-osworld-2-0) |
| 3 | GPT-6 Astra **(new)** | OpenAI | [72.6%](https://openai.com/index/gpt-6-astra/) |
| 4 | Claude Opus 5 **(new)** | Anthropic | [70.6%](https://www-cdn.anthropic.com/ceaf5c7ff2783855203fde8208ec311252dced5b/Claude%20Opus%205%20System%20Card.pdf) |
| 5 | GPT-6.1 Sol **(new)** | OpenAI | [70.5%](https://openai.com/index/introducing-gpt-6-1-sol/) |
[See all 24 entries →](https://leaderboard.steel.dev/leaderboards/osworld-2)
---
### [GAIA](https://leaderboard.steel.dev/leaderboards/gaia)
Model evals / reasoning · Agent scope · 21 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | OPS-Agentic-Search **(new)** | Alibaba Cloud | [92.36%](https://gaia-benchmark-leaderboard.hf.space/ant_test) |
| 1 | openJiuwen-deepagent **(new)** | Suzhou AI Lab / Shuqian Tech | [92.36%](https://gitcode.com/openJiuwen/agent-store/tree/main/community/deepagent) |
| 3 | openJiuwen-deepagent (GPT5/Gemini) | openJiuwen | [91.69%](https://gaia-benchmark-leaderboard.hf.space/ant_test) |
| 4 | Lemon Agent | Lenovo CTO Org | [91.36%](https://github.com/Open-Lemon/LemonAgent) |
| 5 | JoinAI V2.2 | JoinAI-CMCC | [90.7%](https://gaia-benchmark-leaderboard.hf.space/ant_test) |
[See all 21 entries →](https://leaderboard.steel.dev/leaderboards/gaia)
---
### [ClawBench](https://leaderboard.steel.dev/leaderboards/clawbench)
Browser agents · Agent scope · 7 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | Claude Sonnet 4.6 | Anthropic | [33.3%](https://arxiv.org/abs/2604.08523) |
| 2 | GLM-5 **(new)** | Z.ai | [24.2%](https://arxiv.org/abs/2604.08523) |
| 3 | Gemini 3 Flash | Google | [19.0%](https://arxiv.org/abs/2604.08523) |
| 4 | Claude Haiku 4.5 | Anthropic | [18.3%](https://arxiv.org/abs/2604.08523) |
| 5 | GPT-5.4 | OpenAI | [6.5%](https://arxiv.org/abs/2604.08523) |
[See all 7 entries →](https://leaderboard.steel.dev/leaderboards/clawbench)
---
### [HealthAdminBench](https://leaderboard.steel.dev/leaderboards/healthadminbench)
Browser agents · Agent scope · 11 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | Claude Mythos 5 (browser-use) **(new)** | Anthropic | [51.9%](https://www.anthropic.com/claude-fable-5-mythos-5-system-card) |
| 1 | Claude Opus 4.8 (browser-use) **(new)** | Anthropic | [51.9%](https://www.anthropic.com/claude-fable-5-mythos-5-system-card) |
| 3 | Claude Mythos Preview (browser-use) **(new)** | Anthropic | [47.4%](https://www.anthropic.com/claude-fable-5-mythos-5-system-card) |
| 4 | Claude Sonnet 4.6 (browser-use) **(new)** | Anthropic | [45.2%](https://www.anthropic.com/claude-fable-5-mythos-5-system-card) |
| 5 | Claude Opus 4.6 CUA **(new)** | Anthropic | [36.3%](https://arxiv.org/abs/2604.09937) |
[See all 11 entries →](https://leaderboard.steel.dev/leaderboards/healthadminbench)
---
### [Online-Mind2Web](https://leaderboard.steel.dev/leaderboards/online-mind2web)
Browser agents · Agent scope · 25 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | Browser Use Cloud (bu-max) **(new)** | Browser-Use | [97.0%](https://browser-use.com/posts/online-mind2web-benchmark) |
| 2 | GPT-5.4 Native Computer Use | OpenAI | [93.0%](https://openai.com/index/introducing-gpt-5-4/) |
| 3 | ABP + Claude Opus 4.6 | theredsix | [90.53%](https://github.com/theredsix/abp-online-mind2web-results) |
| 4 | TinyFish | TinyFish AI | [90.0%](https://www.tinyfish.ai/blog/mind2web) |
| 5 | CUAWright + GPT-5.5 **(new)** | Microsoft Research / Academic Research | [89.4%](https://arxiv.org/html/2610.04116v1) |
[See all 25 entries →](https://leaderboard.steel.dev/leaderboards/online-mind2web)
---
### [τ-bench](https://leaderboard.steel.dev/leaderboards/tau-bench)
Model evals / reasoning · Model scope · 14 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | Step-3.5-Flash | StepFun | [88.2%](https://arxiv.org/abs/2602.10604) |
| 2 | GLM-4.7 | Z.ai | [87.4%](https://docs.z.ai/guides/llm/glm-4.7) |
| 3 | MiMo-V2-Flash | Xiaomi | [80.3%](https://arxiv.org/abs/2601.02780) |
| 4 | GLM-4.7-Flash | Z.ai | [79.5%](https://inference-docs.cerebras.ai/resources/glm-47-migration) |
| 5 | Claude Fable 5.1 (OpenRouter run) **(new)** | Anthropic | [79.3%](https://openrouter.ai/anthropic/claude-fable-5.1) |
[See all 14 entries →](https://leaderboard.steel.dev/leaderboards/tau-bench)
---
### [AgentBench](https://leaderboard.steel.dev/leaderboards/agentbench)
Model evals / reasoning · Model scope · 10 entries tracked
| Rank | System | Organization | Score |
| ---: | ------ | ------------ | ----- |
| 1 | AgentRL w/ Qwen2.5-32B-Instruct | Tsinghua University | [70.4%](https://arxiv.org/abs/2510.04206) |
| 2 | AgentRL w/ Qwen2.5-14B-Instruct | Tsinghua University | [67.7%](https://arxiv.org/abs/2510.04206) |
| 3 | AgentRL w/ GLM-4-9B-0414 | Tsinghua University | [65.0%](https://arxiv.org/abs/2510.04206) |
| 4 | AgentRL w/ Qwen2.5-7B-Instruct | Tsinghua University | [62.0%](https://arxiv.org/abs/2510.04206) |
| 5 | AgentRL w/ Qwen2.5-3B-Instruct | Tsinghua University | [60.0%](https://arxiv.org/abs/2510.04206) |
[See all 10 entries →](https://leaderboard.steel.dev/leaderboards/agentbench)
<!-- LEADERBOARDS:END -->
## How to read these tables
- **Within-benchmark only.** Scores measure different things on different benchmarks; don't compare a WebVoyager number to a SWE-bench Verified number.
- **Source-linked.** Each score links to the original report — paper, blog post, or official leaderboard. Treat anything self-reported with the appropriate dose of skepticism.
- **Scope matters.** _Agent_ pages reflect full system setups (model + tools + policy). _Model_ pages emphasize base model capability under a stated harness. _Mixed_ pages combine both — read the per-row notes on the site before drawing conclusions.
## Contributing
We welcome new entries, corrections, methodology notes, and new benchmark pages. See
[`CONTRIBUTING.md`](CONTRIBUTING.md) for the evidence standard, JSON schema expectations, ranking
rules, and new-leaderboard checklist.
At minimum, every submitted score needs a public source URL that directly supports the benchmark,
system name, score, and setup notes. If you update leaderboard data, run `npm run update-readme`
before opening a pull request.
## License
MIT