Agent Cache Engineering GPT-OSS-20B - xbench-DS 2510: leaderboard
Agent Cache Engineering (2026): GPT-OSS20B High reasoning appendix, separately named context-management systems;96K Standard/AKV threshold and48K Text Engineering. Same long-horizon task metrics, not serving-replay latency. Success Rate (%).
Metric: Success Rate (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Cache Study GPT-OSS-20B BF16 Standard (High) | 25 | |
| 2 | Cache Study GPT-OSS-20B BF16 AKV (High) | 24 | |
| 3 | Cache Study GPT-OSS-20B BF16 Text Engineering (High) | 19 |
Interactive version: theaggregate.ai/benchmark?slug=agent-cache-engineering-gpt-oss-20b-xbench-ds-2510 · How It Works · Data refreshed daily, snapshot 2026-10-22.