Hebrew LLM - Winograd (0-shot) — leaderboard
Metric: Accuracy (%). Source: huggingface.co. 30 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 96.04 |
| 2 | Gemini 3 Flash (Preview) | 94.6 |
| 3 | Qwen 3.5 122B A10B | 93.17 |
| 4 | GLM-5.1 | 91.37 |
| 5 | GPT-5.1 | 91.01 |
| 6 | Gemini 2.5 Pro | 90.65 |
| 7 | GPT-4o | 89.21 |
| 8 | Claude Sonnet 4.5 | 88.85 |
| 9 | GPT-5 Mini | 88.49 |
| 10 | Qwen 3.5 397B A17B | 86.69 |
| 11 | Gemini 2.5 Flash | 85.97 |
| 12 | Gemma 4 31B (IT) | 85.97 |
| 13 | c4ai-command-a-03-2025 | 84.89 |
| 14 | Llama 3.3 70B Instruct | 83.45 |
| 15 | GLM-4.6 | 83.45 |
Interactive version: aibenchmarks.dev/benchmark?slug=hebrew-llm-winograd-0-shot · How the rankings work · Data refreshed daily, snapshot 2026-07-12.