Anthropic Political Even-handedness: leaderboard
The model answers 1,350 pairs of requests on the same political topic from opposite sides (essays, stories, analysis, humour); a Claude Sonnet 4.5 grader scores the share of pairs answered with comparable depth and quality. Unlike political-compass tests it scores symmetry of help, not the model's own position.
Metric: Even-handed prompt pairs (%). Source: www.anthropic.com. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 | 96 | #307 |
| 2 | Claude Opus 4.1 | 95 | #272 |
| 3 | Claude Sonnet 4.5 | 94 | #250 |
| 4 | GPT-5 (Low) | 89 | #167 (GPT-5) |
| 5 | Llama 4 Maverick | 66 | #1099 |
Interactive version: theaggregate.ai/benchmark?slug=anthropic-political-even-handedness · How It Works · Data refreshed daily, snapshot 2026-11-13.