GameWorld Computer-Use: leaderboard

Computer-use agents playing 34 browser games (170 tasks) via keyboard/mouse control. Measures overall progress percentage using state-verifiable evaluation across runner, arcade, platformer, puzzle, and simulation genres.

Metric: Progress (%). Source: gameworld-project.github.io. Saturation forecast: Around 2033. 8 models tracked.

Top models

#ModelScoreOverall rank
1Median Expert82.6
2Median Human64.1
3Seed 1.839.8#139
4Claude Sonnet 4.638.3#90
5Qwen 3 VL 235B A22B (Thinking)31.4#238 (Qwen 3 VL 235B A22B)
6uitars-1.5-7B31.1#864
7Qwen 3 VL 30B A3B (Thinking)30.8#351 (Qwen 3 VL 30B A3B)

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=gameworld-computer-use · How It Works · Data refreshed daily, snapshot 2026-10-20.