FOCUS - BrowseComp-Plus: leaderboard
FOCUS v1 evaluates task-specific reinforcement-learning and compaction systems, distinct from unmodified backbones. Deontic has 338 held-out cases (150 Airline, 50 SARA-N, and 138 SARA-B, including 71 overlapping Hard cases). BrowseComp uses a fixed corpus with 249 held-out tasks from 830 test tasks; LocBench uses its held-out file-localization protocol. The generation budget is matched at one epoch. Qwen3.5-27B training uses LoRA and is limited to Deontic; its untrained reference results also appear on the other tasks. BrowseComp reports the best held-out checkpoint during training, which is not an unbiased final-checkpoint estimate. Exact source run identities are retained. LocBench GRPO File@1 and File@5 use distinct checkpoints, 1 and 89; FOCUS uses checkpoint 35. The 27B reference remains evaluated even where 27B training was outside the study. Metric: Accuracy (%).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | FOCUS BrowseComp Qwen3.5-9B FOCUS | 55.4 | |
| 2 | FOCUS BrowseComp reference Qwen3.5-27B | 50.6 | |
| 3 | FOCUS BrowseComp Qwen3.5-9B GRPO | 48.2 | |
| 4 | FOCUS BrowseComp Qwen3.5-9B VinePPO | 46.6 | |
| 5 | FOCUS BrowseComp reference Qwen3.5-9B | 29.7 | |
| 6 | FOCUS BrowseComp reference Nemotron 3 Nano 30B A3B | 19.3 | |
| 7 | FOCUS BrowseComp reference gpt-oss-20b (medium) | 18.6 |
Interactive version: theaggregate.ai/benchmark?slug=focus-browsecomp-plus · How It Works · Data refreshed daily, snapshot 2026-10-28.