AI Model Capability Comparison
Computer use
Only models released within the past year are shown. Only benchmarks covering at least 3 of these models are shown. Unknown release dates are excluded; historical data is retained.
ALE-V1 · full · max · Codex
— means no published score for this candidate in this source, not zero.3/11 recent-year models have scores
| No. | Model | Agent engine | Reasoning effort | Split | Tasks run / total tasks | ||
|---|---|---|---|---|---|---|---|
| 1 |  GPT-6 Astragpt-6-astra | Codex | max | 34.21% | 59.26% | full | 151 / 152 |
| 2 | GPT-6 Solgpt-6-sol | Codex | max | 31.58% | 58.57% | full | 151 / 152 |
| 3 | GPT-6 Lunagpt-6-luna | Codex | max | 25% | 50.89% | full | 150 / 152 |
ALE · professional workflows
Source: Agents’ Last Exam ↗ · CC BY 4.0 · Fetched Oct 21, 2026, 06:00 UTC
ALE-V1 · full · max · Codex · Source date unknown
PandaNpc Updated Oct 21, 2026, 06:01 UTC
Source: Epoch AI, CC BY 4.0 · Vals.ai · Artificial Analysis (artificialanalysis.ai)
Column guide
- No.
- Display order only. Sorting changes this number; it is not a published capability rank.
- Model
- The evaluated model name and its provider icon. A dash in a score column means no published score is available here, not a zero.
- Agent engine
- Codex
- Reasoning effort
- max
- Score
- ALE-V1 full split, maximum reasoning effort: percentage of all 152 tasks receiving full credit. Run-task counts may be below 152. Professional workflows allow GUI and command-line tools; Claude Code and Codex use different harnesses. This is not a GUI-only comparison. No upstream snapshot date is published; the fetch time is shown separately. Claude Code and Codex are ranked only within their separate comparison groups. Evaluation methodology ↗
- Partial score
- ALE-V1 full split, maximum reasoning effort: the published average partial task score, converted from 0–1 to a percentage. This is not full-task pass rate. Professional workflows allow GUI and command-line tools; Claude Code and Codex use different harnesses. Run-task counts and the 152-task split are shown. No upstream snapshot date is published; the fetch time is shown separately. Claude Code and Codex are ranked only within their separate comparison groups.
- Split
- Split
- Tasks run / total tasks
- Distinct tasks with a published run / all 152 tasks in the full split. Pass rate uses the full split as its denominator; unrun tasks are not counted as passes.