BABILong (NIAH, v0, 2k): leaderboard
Metric: QA1–5 Mean Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.1 70B Instruct | 78 |
| 2 | GPT-4 Preview (0125) | 77 |
| 3 | Mixtral 8x22B Instruct (v0.1) | 70 |
| 4 | Phi-3-medium-128k-instruct | 67 |
| 5 | Llama 3.1 8B Instruct | 66 |
| 6 | c4ai-command-r-v01 | 63 |
| 7 | Mixtral 8x7B Instruct (v0.1) | 60 |
| 8 | Llama 3 8B Instruct | 58 |
| 9 | Yi-34B-200K | 56 |
| 10 | Phi-3 Mini 128K Instruct | 55 |
| 11 | Mistral 7B Instruct (v0.2) | 52 |
| 12 | Llama 2 7B 32K Instruct | 49 |
| 13 | Yi-9B-200K | 48 |
| 14 | RWKV-6-World-7B | 48 |
| 15 | Llama 2 7B 32K | 45 |
Interactive version: theaggregate.ai/benchmark?slug=babilong-niah-v0-2k · How It Works · Data refreshed daily, snapshot 2026-10-09.