Browser Agent Leaderboards
Track the best-performing AI agents and models across browser agent benchmarks, computer use, research, and coding. Each leaderboard page includes methodology notes, scope labels, and source-linked benchmark results.
Scores
How model sizes and weights are classified
Size classes follow the Artificial Analysis categories: Tiny ≤4B; Small >4–40B; Medium >40–150B; Large >150B total parameters. Active MoE parameters are reported separately.
When parameters are unavailable, we use standard uncached token prices with a 3:1 input/output blend: Small <$0.15/M, Medium $0.15–$1/M, Large >$1/M. “Price proxy” is a pricing category, not a parameter estimate. “Estimated” marks a provisional classification and remains included in filters. “Unclassified” results await model research and match only the all and mixed / uncertain weight filters.
Harnesses inherit the evaluated model stack's class. Multi-model systems use the largest component category as an approximation. Open weights requires released weights for every required model; open harness code alone does not qualify. Unverified backbones and unavailable evaluated checkpoints are marked uncertain.
Results compare evaluated systems, including their tools, compute budgets, and context strategies. Size categories do not guarantee hardware compatibility. Suggest a correction ↗
All benchmarks
WebVoyager
WebVoyager evaluates end-to-end browser agents on 643 tasks across 15 popular real-world websites. Tasks cover search, navigation, form filling, map and travel lookup, shopping, and information retrieval on live pages rather than static snapshots.
| Model | Score |
|---|---|
| browser-control (Fable 5) | 99.19% |
| Alumnium | 98.5% |
| Surfer 2 | 97.1% |
Current leader: browser-control (Fable 5)
BrowseComp
BrowseComp is OpenAI's benchmark for difficult agentic web research: 1,266 short-answer questions where the answer is easy to verify once found but hard to locate without persistent browsing.
| Model | Score |
|---|---|
| GPT-5.6 Sol Ultra | 92.2% |
| GPT-6 Astra | 91.5% |
| Kimi K3 | 91.2% |
Current leader: GPT-5.6 Sol Ultra
DRACO
DRACO (Deep Research Accuracy, Completeness, and Objectivity) is Perplexity's benchmark for deep research systems: 100 open-ended tasks across 10 domains, sourced from real Perplexity Deep Research traffic and graded against expert rubrics on accuracy, completeness, presentation, and citation.
| Model | Score |
|---|---|
| Claude Opus 5 | 88.6% |
| Claude Opus 5.5 | 87.4% |
| Claude Sonnet 5.5 | 87.0% |
Current leader: Claude Opus 5
WebArena
WebArena evaluates browser agents in reproducible, self-hosted websites instead of the open live web. Its 812 tasks cover e-commerce, forum discussion, collaborative software development, content management, maps, and reference lookup.
| Model | Score |
|---|---|
| WebTactix (DeepSeek v3.2) | 74.3% |
| OpAgent | 71.6% |
| ColorBrowserAgent | 71.2% |
Current leader: WebTactix (DeepSeek v3.2)
SWE-bench Verified
SWE-bench Verified is the 500-instance human-reviewed split of SWE-bench, built from real GitHub issues in popular Python repositories. Agents receive an issue and repository state, then generate a patch.
| Model | Score |
|---|---|
| Claude Opus 5 (Vals.ai run) | 97.00% |
| Claude Opus 5 | 96.0% |
| Claude Mythos 5 | 95.5% |
Current leader: Claude Opus 5 (Vals.ai run)
Aider
Aider Polyglot is the coding benchmark behind the Aider pair-programming tool. It tests an LLM on 225 of Exercism's hardest exercises (the ones few models could solve on the earlier single-language benchmark) across C++, Go, Java, JavaScript, Python, and Rust.
| Model | Score |
|---|---|
| gpt-5 (high) | 88.0% |
| gpt-5 (medium) | 86.7% |
| o3-pro (high) | 84.9% |
Current leader: gpt-5 (high)
OSWorld
OSWorld evaluates multimodal computer-use agents in real desktop environments across 369 tasks involving web apps, desktop software, files, and workflows spanning multiple applications.
| Model | Score |
|---|---|
| Qwen3.8-Max | 86.1% |
| Claude Mythos Preview | 85.4% |
| Claude Mythos 5 | 85.0% |
Current leader: Qwen3.8-Max
OSWorld 2.0
OSWorld 2.0 evaluates computer-use agents on 108 long-horizon, end-to-end desktop workflows spanning everyday and professional tasks across self-hosted websites, office suites, files, and multi-application pipelines.
| Model | Score |
|---|---|
| Claude Fable 5.1 | 77.9% |
| Simular Sai | 73.0% |
| GPT-6 Astra | 72.6% |
Current leader: Claude Fable 5.1
GAIA
GAIA evaluates general AI assistants on 466 real-world questions requiring reasoning, web browsing, multimodal understanding, file handling, and tool use.
| Model | Score |
|---|---|
| OPS-Agentic-Search | 92.36% |
| openJiuwen-deepagent | 92.36% |
| openJiuwen-deepagent (GPT5/Gemini) | 91.69% |
Current leader: OPS-Agentic-Search
ClawBench
ClawBench evaluates browser agents on 153 everyday online tasks across 144 live platforms in 15 categories, including purchases, appointments, job applications, and detailed forms.
| Model | Score |
|---|---|
| Claude Sonnet 4.6 | 33.3% |
| GLM-5 | 24.2% |
| Gemini 3 Flash | 19.0% |
Current leader: Claude Sonnet 4.6
HealthAdminBench
HealthAdminBench evaluates computer-use agents on healthcare revenue-cycle work: 135 expert-designed tasks covering prior authorization, appeals and denials management, and durable medical equipment (DME) order processing, executed in four simulated GUI environments (an EHR, two payer portals, and a fax system). It was built by Stanford's Shah Lab with Stanford Healthcare and Kinetic Systems.
| Model | Score |
|---|---|
| Claude Mythos 5 (browser-use) | 51.9% |
| Claude Opus 4.8 (browser-use) | 51.9% |
| Claude Mythos Preview (browser-use) | 47.4% |
Current leader: Claude Mythos 5 (browser-use)
Online-Mind2Web
Online-Mind2Web turns the static Mind2Web idea into a live benchmark of 300 tasks across 136 websites, covering shopping, finance, travel, government, and other consumer workflows.
| Model | Score |
|---|---|
| Browser Use Cloud (bu-max) | 97.0% |
| GPT-5.4 Native Computer Use | 93.0% |
| ABP + Claude Opus 4.6 | 90.53% |
Current leader: Browser Use Cloud (bu-max)
τ-bench
τ-bench evaluates conversational agents in realistic customer-service tasks where the agent must talk to a simulated user, call domain APIs, and follow a policy manual.
| Model | Score |
|---|---|
| Step-3.5-Flash | 88.2% |
| GLM-4.7 | 87.4% |
| MiMo-V2-Flash | 80.3% |
Current leader: Step-3.5-Flash
AgentBench
AgentBench evaluates LLMs as agents across 8 interactive environments, including operating-system tasks, database querying, knowledge graphs, games, lateral-thinking puzzles, ALFWorld, WebShop, and Mind2Web-style browsing.
| Model | Score |
|---|---|
| AgentRL w/ Qwen2.5-32B-Instruct | 70.4% |
| AgentRL w/ Qwen2.5-14B-Instruct | 67.7% |
| AgentRL w/ GLM-4-9B-0414 | 65.0% |
Current leader: AgentRL w/ Qwen2.5-32B-Instruct