Context Arena: leaderboard
Long-context benchmark leaderboard for multi-needle retrieval and reasoning across increasing context lengths.
Metric: Average Score (%). Source: contextbench.ai. Source: benchmarklist.com. Saturation forecast: Around January 2027. 184 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.8 27B (Medium) | 93.98 | #105 (Qwen 3.8 27B) |
| 2 | Qwen 3.8 27B (Low) | 92.79 | #105 (Qwen 3.8 27B) |
| 3 | Qwen 3.8 27B (xHigh) | 90.17 | #105 (Qwen 3.8 27B) |
| 4 | GPT-5.6 Sol (Max) | 85.51 | #19 (GPT-5.6 Sol) |
| 5 | Gemini 3.7 Flash (High) | 83.03 | #23 (Gemini 3.7 Flash) |
| 6 | GPT-5.5 (xHigh) | 79.77 | #31 (GPT-5.5) |
| 7 | Gemini 3.7 Flash (Medium) | 79.56 | #23 (Gemini 3.7 Flash) |
| 8 | GPT-5.5 (High) | 78.96 | #31 (GPT-5.5) |
| 9 | GPT-5.5 (Medium) | 78.59 | #31 (GPT-5.5) |
| 10 | Claude Opus 5 (Max) | 77.33 | #12 (Claude Opus 5) |
| 11 | GPT-5.6 Terra (Max) | 76.59 | #41 (GPT-5.6 Terra) |
| 12 | GPT-5.5 (Low) | 75.03 | #31 (GPT-5.5) |
| 13 | Gemini 3.7 Flash (Low) | 74.48 | #23 (Gemini 3.7 Flash) |
| 14 | Qwen 3.8 Max (xHigh) | 73.72 | #35 (Qwen 3.8 Max) |
| 15 | Grok 4.6 (Medium) | 73.56 | #26 (Grok 4.6) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=context-arena · How It Works · Data refreshed daily, snapshot 2026-10-20.