Benchmarks · Science & research · Agents & computer use
FrontierRSI: Agentic post-train and harness
FrontierRSI's agentic post-training tasks: design RL data that teaches a small model multi-step coding, or improve the harness around a fixed worker model. Scored against the best result any model reached.
9 models · 9 results · data as of 2026-11-18 · frontier-rsi.org ↗
Leaderboard
Opus 5.5 max
All results
9 rows| # | Model | Effort | Score | Harness |
|---|---|---|---|---|
| 1 | Opus 5.5 | max | 19.0% | Claude Code |
| 2 | Gemini 4 Argon | high | 13.5% | Antigravity CLI |
| 3 | Kimi K3.1 | max | 13.4% | Claude Code |
| 4 | GPT-6 Astra | max | 12.5% | Codex |
| 5 | Grok 4.7 | xhigh | 11.8% | Grok Build |
| 6 | Fable 5.2 | max | 8.2% | Claude Code |
| 7 | Gemini 4 Carbon | high | 4.3% | Antigravity CLI |
| 8 | Muse Wonder | max | 2.0% | Muse Code |
| 9 | GPT-6.1 Sol | max | 0.0% | Codex |