BABILong (NIAH, v0, 0k): leaderboard
Metric: QA1–5 Mean Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 30 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4 Preview (0125) | 87 |
| 2 | Llama 3.1 70B Instruct | 85 |
| 3 | Mixtral 8x22B Instruct (v0.1) | 75 |
| 4 | Yi 34B (Base) | 72 |
| 5 | Phi-3-medium-128k-instruct | 72 |
| 6 | mamba-2.8B-hf | 70 |
| 7 | Llama 3.1 8B Instruct | 67 |
| 8 | Mixtral 8x7B Instruct (v0.1) | 65 |
| 9 | Yi-34B-200K | 65 |
| 10 | Llama 3 8B Instruct | 64 |
| 11 | Phi-3 Mini 128K Instruct | 64 |
| 12 | c4ai-command-r-v01 | 64 |
| 13 | Mistral 7B Instruct (v0.2) | 60 |
| 14 | RWKV-6-World-7B | 56 |
| 15 | Yi-9B-200K | 52 |
Interactive version: theaggregate.ai/benchmark?slug=babilong-niah-v0-0k · How It Works · Data refreshed daily, snapshot 2026-10-09.