Intelligence Verdict

The answer

Stage 3: independently tested.

The rule that placed itSomeone other than the maker has tested it: an independent benchmark, a peer-reviewed paper, or a hands-on test by a trusted outlet.

The evidenceOn OSWorld 2.0, 108 long computer workflows that take a person about 1.6 hours each, the best AI agent fully completed only 20.6% of tasks. OSWorld 2.0, arXiv 2606.29537, Jun 28, 2026.

A stage says how far the evidence has got. It is not a rating of the AI and not advice.

1
ClaimedEvidence found
2
LaunchedEvidence found
3
Independently testedPlaced here
4
Tested with real peopleNo evidence yet
5
Regulator-clearedNo approval process exists
6
Widely used with a verified recordNo evidence yet

The evidence

Everything we found.

In ladder order, newest first within each kind. Makers' own claims are labeled as theirs.

Claimed

What the makers claim

OpenAI (company-reported)
“GPT‑5.4 achieves a state-of-the-art 75.0% success rate”
OpenAI, Introducing GPT-5.4 · Mar 5, 2026 · company-reported

Launched

What has launched

ChatGPT agent
ChatGPT agent is available to the public.
OpenAI, Introducing ChatGPT agent (company-reported) · Jul 17, 2025

Tested

What independent tests found

OSWorld 2.0 authors (Yuan, Xie, Yu et al.)
On OSWorld 2.0, 108 long computer workflows that take a person about 1.6 hours each, the best AI agent fully completed only 20.6% of tasks.
“Claude Opus 4.8 with maximum thinking and batched tool calls scores best but still completes only 20.6% of tasks.”
OSWorld 2.0, arXiv 2606.29537 · Jun 28, 2026
OSWorld, reported in the Stanford AI Index
On OSWorld, a test of everyday computer tasks, the best AI scored 66.3% in 2025. People score 72.35%.
Stanford AI Index 2026 · 2026

Not found in our sources yet: tested with real people; how many people use it. No regulator approves AI for this task, so the approval step doesn't apply.

Beside the stage