QwQ 32B ArliAI RpR v1
ArliAI · 2025-04-07 · 32.8B parameters
QwQ-32B-ArliAI-RpR-v1 is ArliAI's Apache-2.0 open-weight fine-tune of Qwen/QwQ-32B for role-play and creative writing with multi-turn reasoning. Its model card lists a 128K-token maximum context while noting 32K as the realistic range.
Benchmark scores
| Benchmark | Score |
|---|---|
| SnakeBench Average Apples | 0.5 |
| SnakeBench Best Apples | 2.0 |
| SnakeBench Rating | 24.6 |
| SnakeBench Total Apples | 14.0 |
| SnakeBench Win Rate | 58.3 |