SkillEvolBench (No-Skill) - Deployment Success: leaderboard
Coding-agent tasks from SkillEvolBench (Lei et al., 2026): 180 verifier-scored tasks in six environments such as code repair, API orchestration and document transformation, each model running in its vendor agent harness (Claude Code, Codex CLI or Gemini CLI). This board takes the no-skill control and scores the 90 deployment-role tasks, which embed a procedure in a broader request, add shortcut traps or require composing skills; success rate in percent, higher is better.
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; no skills or memory, the base-capability control; verifier-scored). Source: arxiv.org. Saturation forecast: Around October 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 38.9 | #158 |
| 2 | Claude Opus 4.6 | 37.8 | #100 |
| 3 | GPT-5.2 Codex | 36.7 | #185 |
| 4 | Gemini 3.1 Pro (Preview) | 35.6 | #91 |
| 5 | Claude Sonnet 4.5 | 35.6 | #244 |
| 6 | Gemini 3 Flash | 35.6 | #159 |
| 7 | GPT-5.3 Codex | 34.4 | #114 |
| 8 | GPT-5.4 | 33.3 | #121 |
| 9 | Claude Opus 4.5 | 32.2 | #152 |
| 10 | Gemini 2.5 Pro | 26.7 | #262 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=skillevolbench-no-skill-deployment-success · How It Works · Data refreshed daily, snapshot 2026-11-12.