Open PL LLM - EQ-Bench: leaderboard
Polish EQ-Bench on the Open PL LLM Leaderboard (SpeakLeash): predicting the intensity of characters' emotions in dialogues. The score is at most 100 and is signed, falling below zero when many answers cannot be parsed; higher is better. Runs not available on the Hugging Face Hub are excluded.
Metric: EQ-Bench Score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 306 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 14B Instruct | 65.19 | #654 |
| 2 | Qwen 2.5 72B Instruct | 64.29 | #469 |
| 3 | Llama 3.1 405B Instruct FP8 | 64 | #400 |
| 4 | Qwen 2 72B Instruct | 63.17 | #579 |
| 5 | Mistral Large 2 (Nov) Instruct (2411) | 63.12 | #480 |
| 6 | Qwen 2.5 32B Instruct | 63.02 | #509 |
| 7 | WizardLM-2 8x22B | 62.81 | #723 |
| 8 | Mistral Large 2 (Jul) | 62.62 | #490 |
| 9 | Llama 3.3 70B Instruct | 61.79 | #541 |
| 10 | Llama 3.1 Nemotron 70B Instruct | 61.55 | #563 |
| 11 | cogito-v1-preview-qwen-14B | 61.27 | |
| 12 | Qwen 3 32B | 61.15 | #449 |
| 13 | DeepSeek V3 (0324) | 60.48 | #349 |
| 14 | QwQ-32B | 60.04 | #422 |
| 15 | Qwen 2.5 32B | 59.81 | #544 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-pl-llm-eq-bench · How It Works · Data refreshed daily, snapshot 2026-11-05.