DeepSynth-Eval (E2E Single-Turn): leaderboard
Metric: Overall checklist score (%; single-pass report from the summarized oracle context; 96 survey-writing tasks, 19,167 atomic checklist requirements; group score min(1, net reward / saturation threshold) with -1 for an incorrect mention, weighted over groups; DeepSeek-V3.2 judge). Source: arxiv.org. Saturation forecast: Around March 2027. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 | 28.3 |
| 2 | Qwen 3 235B A22B 2507 (Thinking) | 24.8 |
| 3 | DeepSeek V3.2 | 23.4 |
| 4 | Qwen 3 30B A3B 2507 (Thinking) | 6.3 |
Interactive version: theaggregate.ai/benchmark?slug=deepsynth-eval-e2e-single-turn · How It Works · Data refreshed daily, snapshot 2026-09-26.