⬢github Python · 237 ★ +229 since we first saw it · pushed 4 h ago · MIT
PostHog/jeeves
Jeeves – Reasoning improves Jev-like decision models
Jeeves is a 9B (Qwen3.5-based) classifier in the Jev family that 'thinks' before deciding, trained with LoRA, a pointer head, SFT, and CISPO. It supports yes/no, multiple-choice, and rating questions via a Jev-compatible API, ships with weights, training code, and datasets, and claims benchmark wins over Kev-9B and Jev.
Why now: It was just released and is being discussed on Hacker News because it outperforms the popular Jev and Kev models on JevBench and held-out test data.
Who it is for: ML engineers building decision/classification pipelines who currently use Jev-like models with a reasoning-model fallback.
Stars over our 34 snapshots: 8 to 237, since 8 h ago.
Where people talked about it
API: https://socialmediatrends-api.osmike.com/v1/repos/PostHog/jeeves