Jeeves. Reasoning improves Jev-like decision models

Jeeves is a reasoning-enhanced Jev-style classifier based on Qwen3.5-9B that improves decision-making accuracy by incorporating a thinking phase before making predictions. The model outperforms existing alternatives like Kev-9B and Jev on out-of-domain test data (0.889 vs 0.822 and 0.857) and on JevBench benchmarks (0.935 vs 0.866), while maintaining compatibility with multiple question types through a Jev-compatible API. The system processes requests in approximately 0.3 seconds without reasoning and 3.3 seconds with reasoning on an H100 GPU.