Tool Use & Agents · sub-raceDominant lead
Long-Horizon Autonomy race
The length of task (in human expert hours) a model completes autonomously half the time.
- Leader
- Fable 5.1
- Margin
- 5.5 hrs
- 23.9% ahead
- Weeks at #1
- 9
- in a row
- Changes
- 5
- in 2026
Warming up the track…
Oct 19, 2026Fable 5.1 leads
2025Apr ’25Jul ’25Oct ’252026Apr ’26Jul ’26Oct ’26
Score history
Top 6 today over the leader's frontier · click to jump to a week
#1 frontierFable 5.1Opus 5.5GPT-6 AstraSonnet 5.5GPT-6.1 SolGemini 4 Argon
Reigns at #1
Longest: 2.5 Pro, 15 weeks (from Mar 2025).
Sibling races
More tool use & agents races
Why this race matters
This number has been doubling every few months. It is the headline metric for agent progress.
How it is scored
Logistic fit of success vs. human task length; reports the 50% point.
Unit: hoursHigher is betterPhoto finish under 0.3 hrsDominant over 2.0 hrs
Data provenance
50% success time-horizon on software and research tasks.
Scores here are illustrative placeholders that mimic real benchmark patterns. Swap in live data through the score provider. Methodology