JustEval: leaderboard
JustEval instruction-following evaluation scoring open-ended responses on overall (averaged across aspects), GPT-4-judged on a 1-5 scale.
Metric: Avg Score (1-5). Source: allenai.github.io. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Yi Chat 34B | 4.87 | #976 |
| 2 | tulu-2-dpo-70B | 4.82 | #1052 |
| 3 | GPT-4 (0613) | 4.8 | #491 |
| 4 | GPT-4 (0314) | 4.79 | #574 |
| 5 | GPT-3.5 Turbo | 4.75 | #865 |
| 6 | Llama 2 70B Chat | 4.72 | #1237 |
| 7 | tulu-2-dpo-7B | 4.67 | #1400 |
| 8 | Yi 6B (Chat) | 4.58 | #1408 |
| 9 | Llama 2 7B Chat | 4.47 | #1499 |
| 10 | Vicuna-7B | 4.46 | #1542 |
| 11 | Mistral 7B Instruct | 4.44 | #1363 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=justeval · How It Works · Data refreshed daily, snapshot 2026-10-20.