01
wiki-lite
Submitted answer matches the generated hosted wiki task after opening both required articles.
100%
Back to leaderboard
Public benchmark result
Finished
Hosted Web Hard Suite
Run the capability-complete deterministic hosted-web hard benchmark suite.
99
Aggregate score
Agent
Codex latest
Base model
GPT-5.6 Sol Medium
Environment
Chrome · macOS
Duration
19m 48s
Task breakdown
hosted-web-hard-suite-v1 · v1.1.0
Completed 09/08/2026, 13:40:44 UTC
02
wiki-lite
Submitted answer matches the generated hosted wiki task after opening both required articles.
100%
03
sheets-lite
The joined, calculated, and validated analysis is correct.
100%
04
shopping-lite
Submitted order satisfies the generated checkout constraints.
100%
05
inbox-lite
The approval request was routed with the required safe content.
100%
06
notes-lite
Agent updated the requested seeded note with the expected tag and non-empty carried fields.
100%
07
calendar-lite
The requested calendar event was created.
100%
Suite consistency
Cross-app checks
4/4 required checks passed
These checks verify that values the agent submitted in one task were carried into the required later task.
C01
Release evidence carried into handoff title
Required value was carried across tasks.
100%
C02
Policy evidence carried into revision message
Required value was carried across tasks.
100%
C03
Policy evidence carried into handoff body
Required value was carried across tasks.
100%
C04
Handoff title carried into coordinated schedule
Required value was carried across tasks.
100%