Fangcun DeepSeek 1017 Report - NCRI reasoning-effort comparison: leaderboard
Fangcun report 2610.35671v1, hosted API evaluations Oct 4–25 2026; Appendix A.1 exact model identities, maximum reasoning effort unless named otherwise, temperature 1/top_p 0.95. GLM 5.3 judge where required. API/format failures excluded by source; common cohort versus successfully completed denominators retained. Primary PDF printed values, not estimated plot heights. Figure 1.6: No-Chain-of-Thought Reasoning Index accuracy on arithmetic, multistep reasoning, and program-output prediction; exact API effort settings preserved as separate model configurations. Off disables CoT; low/max enable different reasoning budgets. Accuracy uses valid responses in each setting, not a guaranteed identical answered subset. Unsupported off modes are absent, never zeros. Metric: Accuracy (%).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (Max) | 99.3 | #3 (GPT-6 Astra) |
| 2 | DeepSeek V4.1 Flash (1017) (Max) | 99.1 | #25 (DeepSeek V4.1 Flash (1017)) |
| 3 | DeepSeek V4 Flash (Max) | 98.9 | #147 (DeepSeek V4 Flash) |
| 4 | GPT-5.6 Sol (Max) | 98.8 | #19 (GPT-5.6 Sol) |
| 5 | GPT-6 Astra (Low) | 98.7 | #3 (GPT-6 Astra) |
| 6 | DeepSeek V4 Pro (Max) | 98.4 | #109 (DeepSeek V4 Pro) |
| 7 | DeepSeek V4.1 Flash (Max) | 98.2 | #36 (DeepSeek V4.1 Flash) |
| 8 | DeepSeek V4.1 Flash (Low) | 96.2 | #36 (DeepSeek V4.1 Flash) |
| 9 | GPT-5.6 Sol (Low) | 95.3 | #19 (GPT-5.6 Sol) |
| 10 | Gemini 3.8 Flash | 94.9 | #24 |
| 11 | Gemini 3.8 Flash (Low) | 94.7 | #24 (Gemini 3.8 Flash) |
| 12 | GPT-5.6 Sol (Non-reasoning) | 67 | #19 (GPT-5.6 Sol) |
| 13 | DeepSeek V4 Pro | 42.3 | #109 |
| 14 | DeepSeek V4.1 Flash (1017) | 38.5 | #25 |
| 15 | DeepSeek V4.1 Flash | 37.2 | #36 |
No result here: #2 Claude Opus 5.5, #6 GPT-6.1 Sol, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=fangcun-deepseek-1017-report-ncri-reasoning-effort-comparison · How It Works · Data refreshed daily, snapshot 2026-10-28.