DiG-bench - Composite Score: leaderboard
DiG-bench composite: expected score across the 70 hidden-rule text games, weighting a win at 0.8 and in-game progress at 0.2, from the paper's agent-strength fit (0–100, humans 100). Rows are model and harness runs, such as the benchmark agent or Claude Code, kept separate; higher is better.
Metric: Expected composite score (%). Source: digbench.ai. Saturation forecast: Around February 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 68.49 | #16 |
| 2 | Claude Fable 5 | 57.33 | #20 |
| 3 | Claude Opus 4.8 | 42.37 | #39 |
| 4 | GPT-5.5 | 27.94 | #33 |
| 5 | GLM-5.2 | 18.57 | #80 |
| 6 | Qwen 3.5 397B A17B | 5.95 | #154 |
| 7 | DeepSeek V4 Pro | 5.68 | #112 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=dig-bench-composite-score · How It Works · Data refreshed daily, snapshot 2026-11-05.