Browser Use Bench v2.1 | 180 tasks | October 29, 2026

MiniMax M3.1 Flash on real browser tasks.

#34 of 40 at medium effort. DeepSeek V4.1 Flash 1017 low scores 23.4 points higher for 1.0× less.

Launch-day provider; model had no catalog price, so cost is repriced from recorded tokens at MiniMax's list price. Many tasks hit the 60-minute limit.

Score50.6

Cost / task13¢

Median time47 min

#	Model	Score	95% range	Cost / task	Median time
34	MiniMax M3.1 Flash medium	50.6	45–56	13¢	47 min
35	MiniMax M3.1 Flash high	49.5	44–55	13¢	48 min
36	MiniMax M3.1 Flash low	48.6	44–53	11¢	44 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.

Compare |full leaderboard

GPT-6.1 Terra79.0
GPT-6.1 Sol81.3
GPT-6 Astra77.9
GPT-6 Sol79.4
GPT-5.6 Luna69.6
GPT-6 Luna69.2
Opus 5.575.0
Opus 566.8
Sonnet 5.572.7
Fable 5.172.0
Haiku 5.569.5
Grok 4.773.3
DeepSeek V4.1 Flash70.9
DeepSeek V4.1 Flash 101775.0
GLM 5.3 FlashX59.7
Kimi K3.1 Pro76.7

Run your tasks in the cloud