AgenticRL-Bench Open-ended Head-to-head Elo: leaderboard

AgenticRL-Bench Open-ended Head-to-head Elo — 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: Head-to-head Elo. Source: https://agenticrl-bench.env-lab.ai/leaderboard

Metric: Head-to-head Elo. Source: agenticrl-bench.env-lab.ai. Saturation forecast: Rough model projection: around 2026. 7 models tracked. Updated 16 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Fable 5.2(Max) (AgenticRL run: harness=Claude Code)1248#1
2GPT-6 Astra(Max) (AgenticRL run: harness=Codex)1180#6
3Claude Opus 5.5(Max) (AgenticRL run: harness=Claude Code)1036#5
4Gemini 4 Argon(High) (AgenticRL run: harness=Antigravity)1008#12
5Gemini 4 Carbon(High) (AgenticRL run: harness=Antigravity)911#2
6Grok 4.7(Xhigh) (AgenticRL run: harness=Grok Build)826#47
7Muse Wonder Full 1.0(Xhigh) (AgenticRL run: harness=Muse Code)791

Interactive version: theaggregate.ai/benchmark?slug=agenticrl-bench-open-ended-head-to-head-elo · How It Works · Data refreshed daily, snapshot 2026-11-16.