ReScopeBench - Update Coverage: leaderboard
A long-lived tool agent acquires and updates experience across 18 evolving environments, with 352 tasks and 26 hidden changes per condition. Metrics distinguish task completion from judged reference knowledge coverage. 18 versioned environments; 352 tasks and 26 changes per condition; Hermes-based paired native/reset streams; maximum supported reasoning effort where configurable; temperature 1/top-p .95 where supported; 100 foreground iterations per task; 65,536 output tokens per response; GPT-6 Astra medium judge for experience claims; equal environment weighting. Metric: Immediate reference revision coverage (%).
Metric: Immediate reference revision coverage (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Argon (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 37.63 | |
| 2 | Gemini 3.8 Flash (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 33.53 | |
| 3 | DeepSeek V4.1 Flash (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 33.41 | |
| 4 | Claude Opus 5.5 (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 33.27 | |
| 5 | GLM-5.3 (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 33 | |
| 6 | DeepSeek V4 Pro (0813) (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 32.44 | |
| 7 | GLM-5.3 Flash (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 32.07 | |
| 8 | Qwen 3.8 Max (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 31.1 | |
| 9 | Seed 2.1 Pro (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 29.75 | |
| 10 | GPT-6 Astra (Benchmark Sweep Run: source=ReScopeBench; harness=Hermes-based; effort=maximum-supported; exact-API-snapshot=unreported) | 29.37 |
Interactive version: theaggregate.ai/benchmark?slug=rescopebench-update-coverage · How It Works · Data refreshed daily, snapshot 2026-11-06.