Tool Use & Agents · sub-raceClear lead
Computer Use race
Operating real desktop apps through screenshots, clicks and keystrokes.
- Leader
- GPT-6 Astra
- Margin
- 1.7 pts
- 2.2% ahead
- Weeks at #1
- 7
- in a row
- Changes
- 5
- in 2026
Warming up the track…
Oct 19, 2026GPT-6 Astra leads
2025Apr ’25Jul ’25Oct ’252026Apr ’26Jul ’26Oct ’26
Score history
Top 6 today over the leader's frontier · click to jump to a week
#1 frontierGPT-6 AstraOpus 5.5Haiku 5.5Sonnet 5.5Fable 5.1Muse Wonder
Reigns at #1
Longest: Opus 4.5, 13 weeks (from Nov 2025).
Sibling races
More tool use & agents races
Why this race matters
Computer use lets AI work with any software, no API required.
How it is scored
Task success rate within a step budget.
Unit: percentHigher is betterPhoto finish under 0.5 ptsDominant over 3.0 pts
Data provenance
Open-ended tasks across real operating systems.
Scores here are illustrative placeholders that mimic real benchmark patterns. Swap in live data through the score provider. Methodology