Ranked 2 of 9 on the combined score
GPT 6.1 Sol
Codex, max reasoning. 400 timed runs on 200 tasks from 18 benchmarks.
Duration following
Does it work for as long as it is asked? Each task was asked for a short and a long duration: a quarter and four times its own forecast of each task (on PaperBench and PostTrainBench, of its runtime with no time asked).
| Ask | Timing error | Where runs ended |
|---|---|---|
| Every timed run400 runs | 1.15× | |
| Short ask200 runs | 1.21× | |
| Long ask200 runs | 1.09× |
It ran long on short asks (median 1.07× at the short ask) and was within 5% of the time asked on long ones (median 1.01× at the long ask). See every run
Every Sol 6.1 run
Each row is one run, with a mark on the chart. Hover a row to find its mark; click a mark to find its row.
400 runs
| Loading 400 runs. | |||
No runs match these filters.
Forecasting
Before each task, it forecast how long the task would take. Each forecast is set against its own runtime on that task: its run with no time asked.
- Typical miss
- 2.85×
- Geometric mean of the factor between forecast and runtime
- Most forecasts
- 90% too high
- On 200 tasks
- Median forecast
- 25 min
- Median runtime
- 9 min
Retrospection
After its runs had finished, it was asked how long they took, with less time information at each step. Each figure is its typical miss against the real runtime, two answers per condition.
- Oracle
- 1.00×
- Native
- 1.46×
- Context-only
- 1.51×
- Replay
- 1.86×
- Scrubbed
- 1.91×
Oracle adds an elapsed-time tool. Native and Context-only continue the finished session, with tools on or off. Replay sends a rebuilt transcript to the model's API, and Scrubbed does the same without timestamps or other time cues. On 23 tasks.
Across the 18 benchmarks
Timing on its timed runs, and how good the work was. Score is the mean of its graded runs out of 100, from its runs with no time asked; the benchmark score is the mean of these, each benchmark counting the same.
| GPQA Diamond24 runs | 1.66× | 17% | 100.012 graded |
| OSWorld36 runs | 1.62× | 39% | 75.417 graded |
| Humanity's Last Exam40 runs | 1.61× | 43% | 55.020 graded |
| DeepSWE28 runs | 1.17× | 86% | 85.714 graded |
| AssistantBench16 runs | 1.16× | 50% | 6.28 graded |
| PaperBench8 runs | 1.13×(few runs) | 75% | Not scored |
| TUA-Bench20 runs | 1.10× | 55% | 75.110 graded |
| AppWorld28 runs | 1.09× | 50% | 57.114 graded |
| BrowseComp24 runs | 1.08× | 63% | 83.312 graded |
| Agents' Last Exam24 runs | 1.06× | 71% | 64.512 graded |
| YC-Bench8 runs | 1.05×(few runs) | 63% | Own units |
| Terminal-Bench 4.032 runs | 1.04× | 81% | 62.516 graded |
| CORE-Bench24 runs | 1.04× | 75% | 100.012 graded |
| PPTArena20 runs | 1.04× | 65% | 45.010 graded |
| METR public tasks4 runs | 1.02×(few runs) | 75% | Too few graded |
| Sakana ALE-Bench20 runs | 1.02× | 95% | Own units |
| ProgramBench36 runs | 1.01× | 100% | 71.118 graded |
| PostTrainBench8 runs | 1.01×(few runs) | 100% | Not scored |
No runs match these filters.
Highest timing error first. ALE-Bench and YC-Bench keep their own units and stay out of the score, as do PaperBench and PostTrainBench.