NextAct - K10 Native Reranker Overall: leaderboard
NextAct pinned preprint Table 3: same 1,500 forecast points, 50-event context and best-of-three generation as embedding results, scored with Qwen3-Reranker-8B. Gemini is the Preview endpoint. Codex Retrieval supplies DeepSeek additional history; the Native condition has no retrieval. Small-model reranker cells are unreported and remain missing. Metric: Chance-normalized reranker score.
Metric: Chance-normalized reranker score. Source: kkkkawayi.github.io. Saturation forecast: Around November 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol | 0.08 | #19 |
| 2 | Gemini 3.1 Pro (Preview) | 0.07 | #61 |
| 3 | Claude Opus 4.6 | 0.07 | #65 |
| 4 | DeepSeek V4 Flash | 0.06 | #145 |
No result here: #5 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=nextact-k10-native-reranker-overall · How It Works · Data refreshed daily, snapshot 2026-10-26.