AgenticRL-Bench GLM-4.7-Flash Sync Step Speedup: leaderboard
AgenticRL-Bench GLM-4.7-Flash Sync Step Speedup — 19verifiable tasks×3attempts, headless one-shot8h, pass@1 solved/57. Three open-ended4×H200 tasks10–16h, means of3counted attempts in their native units. Elo is native BT within-task comparisons, not accuracy. Exact model/harness/effort retained; no best-harness pooling. Metric: Correct training-step speedup (x). Source: https://agenticrl-bench.env-lab.ai/leaderboard
Metric: Correct training-step speedup (x). Source: agenticrl-bench.env-lab.ai. Saturation forecast: Rough model projection: around 2026. 7 models tracked. Updated 16 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-6 Astra | (Max) (AgenticRL run: harness=Codex) | 1.56 | #6 |
| 2 | Claude Fable 5.2 | (Max) (AgenticRL run: harness=Claude Code) | 1.42 | #1 |
| 3 | Claude Opus 5.5 | (Max) (AgenticRL run: harness=Claude Code) | 1.37 | #5 |
| 4 | Gemini 4 Argon | (High) (AgenticRL run: harness=Antigravity) | 1.3 | #12 |
| 5 | Muse Wonder Full 1.0 | (Xhigh) (AgenticRL run: harness=Muse Code) | 1.19 | |
| 6 | Grok 4.7 | (Xhigh) (AgenticRL run: harness=Grok Build) | 1.17 | #47 |
| 7 | Gemini 4 Carbon | (High) (AgenticRL run: harness=Antigravity) | 0.45 | #2 |
Interactive version: theaggregate.ai/benchmark?slug=agenticrl-bench-glm-4-7-flash-sync-step-speedup · How It Works · Data refreshed daily, snapshot 2026-11-16.