Long-context Evaluation on RULER ultra-long context official (Context Length Sweep)
96Accuracy (128K)Qwen3-Next-80B-A3B-Instruct
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-Next-80B-A3B-InstructParameters=80B2026.02 | 96 | 86.9 | 80.3 | — | |
| Qwen3-235B-A22B-Instruct-2507Parameters=235B, Version=25072026.02 | 93.9 | 90.9 | 84.5 | — | |
| MiniCPM-SALAParameters=9B2026.02 | 89.4 | 87.1 | 86.3 | 81.6 | |
| Qwen3-30B-A3B-Instruct-2507Parameters=30B, Version=25072026.02 | 89.1 | 78.4 | 72.8 | — |