DiagnosticIQ Pro: leaderboard
Metric: Macro accuracy (%) on DiagnosticIQ Pro, the 10-option variant of each question (mean of the 16 per-asset accuracies), zero-shot multiple choice, temperature 0, at most 4,096 output tokens; higher is better. Source: arxiv.org. Saturation forecast: Around September 2027. 30 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 | 59.81 |
| 2 | GPT-5.4 | 59.21 |
| 3 | Gemini 3.1 Pro (Preview) | 57.74 |
| 4 | Claude 3.7 Sonnet | 56.63 |
| 5 | Qwen 3.6 35B A3B | 55.61 |
| 6 | Gemma 4 26B A4B | 47.65 |
| 7 | Claude Sonnet 4.6 | 47.21 |
| 8 | Gemma 4 31B | 45.18 |
| 9 | Llama 4 Maverick | 42.65 |
| 10 | Granite 3.3 8B Instruct | 42.39 |
| 11 | DeepSeek V3 | 41.38 |
| 12 | GPT-5 | 40.69 |
| 13 | Llama 3.1 405B | 38.82 |
| 14 | Gemini 2.5 Pro | 37.51 |
| 15 | Llama 3.3 70B Instruct | 36.56 |
Interactive version: theaggregate.ai/benchmark?slug=diagnosticiq-pro · How It Works · Data refreshed daily, snapshot 2026-10-07.