LiveBench Browser Use

Multi-step tasks in working web apps, done in a real browser by screenshot. Score is the share of tasks passed.

ModelsModels with a result.49
Top resultBest result on this benchmark.100.0%GPT-6.1 Terra
Top-3 spreadPoints from first to third.4.0 pts
UpdatedDate the source changed the results.14 Oct 2026

Result and cost

49 of 49 results

Results

49 results
#ModelResultThe score from the source.CostUS dollars to run one task.
03Claude Opus 4.8max effortAbove fitFar above this model's index.96.0%$8.87
42Kimi K3max effort16.0%$0.382
42Inkling16.0%$2.40