NeedleBench — leaderboard
NeedleBench: Measures long-context retrieval, needle finding, summarization, factual grounding, or retrieval-augmented generation quality.
Metric: Overall 128K score (self-reported). Source: benchmarklist.com. Status: saturation imminent. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2.5 72B Instruct | 81.02 |
| 2 | Gemma 3 27B | 80.38 |
| 3 | Qwen 2.5 32B | 78.25 |
| 4 | Gemma 3 12B | 75.31 |
| 5 | Qwen 2.5 14B | 73.96 |
| 6 | Llama 3.1 70B | 72.37 |
| 7 | Llama 3.1 8B | 70.98 |
| 8 | Qwen 2.5 7B | 70.75 |
| 9 | GLM-4 9B Chat | 66.51 |
| 10 | Gemma 3 4B | 64.42 |
Interactive version: theaggregate.ai/benchmark?slug=needlebench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.