Into the Unknown (strict) - Completeness: leaderboard
AI agents extend 15 published scientific studies and write follow-up papers. This score is the percentage of papers passing every check for completing the requested research.
Metric: Papers passing Completeness (%), requiring all counted nodes in each applicable gate; 15 scientist-proposed open research follow-ups; GPT-6 Astra max via Codex SDK, full workspace, separated rubrics; per-node majority of four judgments (3/4 required); source agent SDK/CLI and effort retained; higher is better. Explicit zero results are retained.. Source: prometheus-eval.github.io. Source: raw.githubusercontent.com. Saturation forecast: Around 2031. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (Max) | 6.67 | #5 (GPT-6 Astra) |
| 2 | Kimi K3 | 0 | #30 |
| 3 | GLM-5.3 | 0 | #40 |
| 4 | GPT-5.6 Sol (Max) | 0 | #18 (GPT-5.6 Sol) |
| 5 | Claude Opus 5 (Max) | 0 | #12 (Claude Opus 5) |
| 6 | Claude Opus 4.8 (Max) | 0 | #38 (Claude Opus 4.8) |
| 7 | DeepSeek V4 Pro (0813) | 0 | #53 |
| 8 | Claude Fable 5 (Max) | 0 | #15 (Claude Fable 5) |
| 9 | Claude Opus 5.5 (Max) | 0 | #3 (Claude Opus 5.5) |
No result here: #7 Gemini 4 Argon, #8 GPT-6.1 Sol, #9 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=into-the-unknown-strict-completeness · How It Works · Data refreshed daily, snapshot 2026-10-16.