BROWSER USE - Browser Use Agents: give Browser Use a task and receive completed work. API V4 is current for new integrations. - Browser Infrastructure: connect your agent or automation to managed browsers through SDK, REST, or CDP. Starts at $0.02/browser-hour. - Developer tools: Open Source, Browser Harness, SDK, and MCP support the two products above. [Developer Index](https://browser-use.com/index.md) [Product Map](https://browser-use.com/llms.txt) [Full Product Context](https://browser-use.com/llms-full.txt) [Pricing](https://browser-use.com/pricing.md) [Cloud Docs](https://docs.browser-use.com/cloud/quickstart) [Open Source Docs](https://docs.browser-use.com/open-source/introduction) --- # Agent benchmarks: best LLM for browser agents Browser Use Bench v2.1 | 180 tasks | October 7, 2026 ## Agent benchmarks. GPT-6.1 Sol medium scores highest at 79.6. Browser Use's own agent scores 72.9 at 11¢ a task, level with Claude Sonnet 5.5 high at 14× less. [Agents](https://browser-use.com/benchmarks/agents)[Browsers](https://browser-use.com/benchmarks/browsers) - OpenAI - Anthropic - xAI - DeepSeek - Z.ai - Browser Use All reasoning efforts # Model Score 95% range Cost / task Median time 1 [GPT-6.1 Sol medium](https://browser-use.com/benchmarks/models/gpt-6-1-sol) 79.6 76–83 $1.30 8 min 2 [GPT-6 Astra medium](https://browser-use.com/benchmarks/models/gpt-6-astra) 77.9 74–82 $10.77 9 min 3 [Claude Opus 5.5 high](https://browser-use.com/benchmarks/models/claude-opus-5-5) 75.0 71–79 $2.35 12 min 4 [GPT-6 Sol medium](https://browser-use.com/benchmarks/models/gpt-6-sol) 74.4 71–78 $1.02 6 min 5 [Grok 4.7 high](https://browser-use.com/benchmarks/models/grok-4-7) 73.3 70–77 $3.46 18 min 6 Browser Use (BU fast) 72.9 69–77 11¢ 9 min 7 [Claude Sonnet 5.5 high](https://browser-use.com/benchmarks/models/claude-sonnet-5-5) 72.7 69–77 $1.53 9 min 8 [Claude Fable 5.1 high](https://browser-use.com/benchmarks/models/claude-fable-5-1) 72.0 68–76 $5.19 17 min 9 [DeepSeek V4.1 Flash max](https://browser-use.com/benchmarks/models/deepseek-v4-1-flash) 70.9 67–75 18¢ 18 min 10 [GPT-5.6 Luna xhigh](https://browser-use.com/benchmarks/models/gpt-5-6-luna) 69.6 66–73 28¢ 15 min 11 [Claude Haiku 5.5 xhigh](https://browser-use.com/benchmarks/models/claude-haiku-5-5) 69.5 65–73 46¢ 14 min 12 [GPT-6 Luna xhigh](https://browser-use.com/benchmarks/models/gpt-6-luna) 69.0 65–73 16¢ 11 min 13 [Claude Opus 5](https://browser-use.com/benchmarks/models/claude-opus-5) 66.8 62–71 $4.72 19 min 14 [Claude Haiku 5.5 max](https://browser-use.com/benchmarks/models/claude-haiku-5-5) 65.3 61–70 $1.07 21 min 15 [Claude Haiku 5.5 high](https://browser-use.com/benchmarks/models/claude-haiku-5-5) 64.3 60–68 23¢ 9 min 16 [GPT-6 Luna medium](https://browser-use.com/benchmarks/models/gpt-6-luna) 64.3 60–68 7.7¢ 8 min 17 [GLM 5.3 FlashX max](https://browser-use.com/benchmarks/models/glm-5-3-flashx) 59.7 55–64 55¢ 14 min 18 [Claude Haiku 5.5 medium](https://browser-use.com/benchmarks/models/claude-haiku-5-5) 53.4 50–57 9.2¢ 5 min 19 [Claude Haiku 5.5 low](https://browser-use.com/benchmarks/models/claude-haiku-5-5) 45.6 42–50 2.7¢ 2 min 180 long tasks on live websites, the same for every model, run in the open-source BrowserCode harness (Browser Use's own agent runs in ours). Mean weighted rubric score per task, 0-100. Estimated model cost per task in USD. Excludes search, judge, browser and runner costs. Runner and encrypted tasks: [github.com/browser-use/benchmark](https://github.com/browser-use/benchmark). Each model - [GPT-6.1 Sol79.6](https://browser-use.com/benchmarks/models/gpt-6-1-sol) - [GPT-6 Astra77.9](https://browser-use.com/benchmarks/models/gpt-6-astra) - [GPT-6 Sol74.4](https://browser-use.com/benchmarks/models/gpt-6-sol) - [GPT-5.6 Luna69.6](https://browser-use.com/benchmarks/models/gpt-5-6-luna) - [GPT-6 Luna69.0](https://browser-use.com/benchmarks/models/gpt-6-luna) - [Opus 5.575.0](https://browser-use.com/benchmarks/models/claude-opus-5-5) - [Opus 566.8](https://browser-use.com/benchmarks/models/claude-opus-5) - [Sonnet 5.572.7](https://browser-use.com/benchmarks/models/claude-sonnet-5-5) - [Fable 5.172.0](https://browser-use.com/benchmarks/models/claude-fable-5-1) - [Haiku 5.569.5](https://browser-use.com/benchmarks/models/claude-haiku-5-5) - [Grok 4.773.3](https://browser-use.com/benchmarks/models/grok-4-7) - [DeepSeek V4.1 Flash70.9](https://browser-use.com/benchmarks/models/deepseek-v4-1-flash) - [GLM 5.3 FlashX59.7](https://browser-use.com/benchmarks/models/glm-5-3-flashx) [Run your tasks in the cloud](https://cloud.browser-use.com/v4?utm_source=browser-use&utm_medium=benchmark-models) ›Earlier benchmarks: Internal Bench Hard, OnlineMind2Web, BU Bench V1 Internal Bench HardOnlineMind2WebBU Bench V1 ### Internal Bench Hard Browser Use: 82% of tasks solved at 17¢ each. That is 20 points better than Opus 5, which costs 20× more per solved task. Toolstrict accuracyCost Browser Use82%$0.17 Opus 562%$3.40 Gemini 3.1 Pro59%$2.20 Sonnet 559%$1.55 GPT-5.652%$1.10 Gemini 3.6 Flash46%$0.62 GPT-537%$0.44 Toolstrict accuracyCost Browser Use82%$0.17 Opus 562%$3.40 Gemini 3.1 Pro59%$2.20 Sonnet 559%$1.55 GPT-5.652%$1.10 Gemini 3.6 Flash46%$0.62 GPT-537%$0.44 Internal Bench Hard · updated 2026-08-01 · [all benchmarks](https://browser-use.com/benchmarks/agents)