CUA-bench (Vals AI) leaderboard
- GPT-6 Astra leads
- 19.2%
- models measured
- 8
- lead over #2
- +5.2 pts
- median
- 8.7%
Leader on CUA-bench (Vals AI) · read October 22
On October 22, 2026, 8 of our catalog's models had a CUA-bench (Vals AI) result, and OpenAI GPT-6 Astra led with 19.2%, ahead of Anthropic Claude Opus 5.5 at 14.0%.
Computer-use agents playing commercial video games with a keyboard and mouse, on games held out from the agents. The score as Vals AI publishes it.
Role in the index
CUA-bench (Vals AI) is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAI | 19.2% | 85.1 | $50.00 | |
| 2 | Claude Opus 5.5Anthropic | 14.0% | 88.8 | $20.00 | |
| 3 | Claude Fable 5.1Anthropic | 13.2% | 84.2 | $50.00 | |
| 4 | Claude Opus 5Anthropic | 9.0% | 83.0 | $25.00 | |
| 5 | GPT-5.6 SolOpenAI | 8.3% | 73.5 | $20.00 | |
| 6 | Muse WonderMeta | 7.2% | 71.9 | $20.00 | |
| 7 | Gemini 4 ArgonGoogle | 4.8% | 81.0 | $10.00 | |
| 8 | Gemini 3.8 FlashGoogle | 4.2% | 61.6 | $3.75 |
Only models in our catalog are listed, and each appears once with its best published run. 137 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.