Confabulation Leaderboard (Lechmazur) — leaderboard
Metric: Confabulation rate % (lower is better). Source: lechmazur.github.io. 64 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4 (Thinking 16K) | 2.48 |
| 2 | Claude Sonnet 4 (Thinking 16K) | 2.48 |
| 3 | Claude Opus 4 (Non-reasoning) | 2.97 |
| 4 | Gemini 2.5 Pro | 3.96 |
| 5 | Grok 4 | 3.96 |
| 6 | Gemini 2.5 Pro (Preview 03-25) | 3.96 |
| 7 | Claude Opus 4.1 (Non-reasoning) | 3.96 |
| 8 | Claude Sonnet 4 (Non-reasoning) | 5.45 |
| 9 | Gemini 2.5 Pro (Preview 05-06) | 5.94 |
| 10 | Grok 3 Mini (High) | 6.93 |
| 11 | GLM-4.5 | 7.92 |
| 12 | O1 | 10.89 |
| 13 | Grok 3 Mini (Low) | 10.89 |
| 14 | GPT-4.5 (Preview) | 11.88 |
| 15 | Claude 3.5 Sonnet (20241022) | 12.87 |
Interactive version: theaggregate.ai/benchmark?slug=confabulation-leaderboard-lechmazur · How the rankings work · Data refreshed daily, snapshot 2026-07-22.