Instruction Following on EQ-Bench Longform (ID prompts) (45 prompts)
1,911Elo ScoreGPT-5.4
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5.4Judge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,911 | |
| Claude Opus 4.6Judge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,774 | |
| GPT-5.4-miniJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,728 | |
| POLARIS–9BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,658 | |
| Gemini 3.1 ProJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,619 | |
| Gemini 3 FlashJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,612 | |
| GPT-5.4-nanoJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,563 | |
| Gemma 4 31BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,505 | |
| Qwen3.5-27BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,494 | |
| Gemma 4 26B-A4BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,451 | |
| Qwen3.5-9BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,343 | |
| Gemma 4 E4BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,301 | |
| Ministral-3-8BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,214 | |
| DeepSeek-R1-distill-Qwen-14BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,033 | |
| LongWriter-Zero-32BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 1,011 | |
| LongWriter-Llama3.1-8BJudge model=Gemini 3 Flash, Judge dimensions=11 dims + length, Comparison protocol=dual-position, Prompt count=452026.06 | 800 |