K-BrowseComp - Calibration Error: leaderboard
Metric: Expected calibration error (%) of the stated confidence on K-BrowseComp-Verified (five equal-width bins, weighted gap between mean confidence and accuracy); lower is better. Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Pro | 17.72 |
| 2 | Gemma 4 31B (IT) | 23.66 |
| 3 | K-EXAONE | 24.09 |
| 4 | GLM-5.1 | 27.07 |
| 5 | GPT-5.5 | 31.86 |
| 6 | GPT-5.4 Mini | 37.88 |
| 7 | Qwen 3.6 35B A3B | 47.89 |
| 8 | A.X 4.0 | 47.89 |
| 9 | Gemini 3.1 Flash Lite | 56.55 |
| 10 | HyperCLOVA X SEED Think (32B) | 77.37 |
Interactive version: theaggregate.ai/benchmark?slug=k-browsecomp-calibration-error · How It Works · Data refreshed daily, snapshot 2026-09-29.