Conceptual Reasoning Index - Decision Theory (DTBench): leaderboard
The decision-theory component of the CRI, using the 407 capability questions from a handcrafted 537-item multiple-choice set on Newcomb-like problems, such as an agent interacting with near-copies of itself. Answers are marked against an expert-validated key, and accuracy is rescaled so the 40 percent random-guessing level maps to 0 and 100 percent to 100. Higher is better.
Metric: Chance-Corrected Score (0-100). Source: conceptualreasoning.ai. Saturation forecast: Estimated already saturated. 267 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 98.22 | #3 |
| 2 | Claude Fable 5 | 97.33 | #16 |
| 3 | Claude Opus 5 (High) | 96.45 | #12 (Claude Opus 5) |
| 4 | Claude Opus 5 (xHigh) | 96.45 | #12 (Claude Opus 5) |
| 5 | Claude Opus 5 | 96 | #12 |
| 6 | Claude Fable 5.1 | 96 | #9 |
| 7 | Claude Fable 5.1 (xHigh) | 96 | #9 (Claude Fable 5.1) |
| 8 | GPT-6 Sol | 95.55 | #14 |
| 9 | Grok 4.6 | 95.55 | #26 |
| 10 | Gemini 4 Argon | 95.55 | #8 |
| 11 | Claude Sonnet 5.5 | 95.55 | #11 |
| 12 | GPT-6 Astra (xHigh) | 95.55 | #6 (GPT-6 Astra) |
| 13 | Claude Opus 5.5 (xHigh) | 95.55 | #3 (Claude Opus 5.5) |
| 14 | Claude Fable 5 (xHigh) | 95.55 | #16 (Claude Fable 5) |
| 15 | Gemini 4 Argon (Medium) | 95.55 | #8 (Gemini 4 Argon) |
No result here: #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=conceptual-reasoning-index-decision-theory-dtbench · How It Works · Data refreshed daily, snapshot 2026-10-20.