EKI Estonian LLM - Propaganda Resistance: leaderboard
EKI propaganda-resistance benchmark: 75 questions on Kremlin strategic narratives, asked neutrally, with a biased premise or maliciously, each in Estonian, English and Russian; a Claude Opus 4.5 judge rates answers 1-5. Geometric mean rescaled to 0-100, higher means more resistant.
Metric: Geometric mean of 1-5 judge ratings, rescaled to 0-100. Source: moodupuu.eki.ee. Saturation forecast: Around February 2027. 100 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 97.59 | #16 |
| 2 | Grok 4.7 | 97.38 | #26 |
| 3 | Claude Fable 5.1 | 97.05 | #10 |
| 4 | GLM-5.3 | 95.7 | #44 |
| 5 | Grok 4.6 | 95.7 | #29 |
| 6 | Grok 4.5 | 95.24 | #54 |
| 7 | Claude Fable 5 | 95.23 | #20 |
| 8 | Claude Opus 4.7 | 94.88 | #51 |
| 9 | GLM-5.3 Flash | 94.38 | #77 |
| 10 | Claude Sonnet 5.5 | 93.42 | #14 |
| 11 | Claude Opus 5.5 | 93.24 | #3 |
| 12 | GPT-5.6 Terra | 92.96 | #40 |
| 13 | GPT-5.6 Sol | 92.79 | #22 |
| 14 | Kimi K3 | 92.77 | #35 |
| 15 | Claude Opus 4.8 | 92.73 | #39 |
No result here: #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=eki-estonian-llm-propaganda-resistance · How It Works · Data refreshed daily, snapshot 2026-11-05.