SkillEvolBench (SelfGen-Zero-Shot) - Deployment Success: leaderboard
Metric: Frozen deployment success rate (%; ESR, share of the 90 context-shift, adversarial and composition tasks, 30 of each, solved with the environment's frozen skill library; fixed skills the model generates from task metadata before any attempt; verifier-scored). Source: arxiv.org. Saturation forecast: Around September 2027. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 (Claude Code) | 36.7 |
| 2 | Claude Opus 4.6 (Claude Code) | 32.2 |
Interactive version: theaggregate.ai/benchmark?slug=skillevolbench-selfgen-zero-shot-deployment-success · How It Works · Data refreshed daily, snapshot 2026-09-26.