Jeeves
PostHog's open-source 9B Jev-style decision classifier (Qwen3.5-9B, LoRA) that reasons before it decides via a diffusion drafter, beating Jev on JevBench's public tiers.
Why it matters
First open-source proof that a small decision classifier can think before it decides and beat the Jev baseline it replaces, with a confidence gate to skip reasoning on easy calls.
What you could build with it
A customer-support tooling startup could route ticket triage through Jeeves: fast 0.3-second yes/no decisions for clear-cut cases, with the confidence gate reserving the slower reasoning pass only for ambiguous tickets.
Does it hold up?
Weights on Hugging Face with full training code and data; runs on CUDA or Apple Silicon, 0.3s per request without reasoning vs 3.3s median with it. The accuracy/latency tradeoff is honestly documented in the repo and an independent review.
Built with Jeeves
- Jeeves Lets a 0.3-Second Classifier Think. Its p90 Reaches 17.1 Seconds.Mark Huang (independent review) · Hands-on analysis of Jeeves' accuracy-vs-latency tradeoff on 325 dev questions: reasoning lifted accuracy 0.775 to 0.825 but pushed p90 latency to 17.1 seconds, recommending the confidence gate for uncertain decisions only.
- jaredpalmer/kevgithub · 8,441 ★ · Jev-like family of decision models built on top of Qwen3.5/3.8 that inspired Jeeves and serves as its published benchmark baseline (Kev-9B).
Learn more
First spotted on github: source.
More AI models
ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71Mistral Forge: enterprise platform for training and continuously improving proprietary modelsMistral AI launched Forge (Sep 28) — a full-lifecycle model training platform (pre-training, SFT, DPO/ODPO, RL…models · JEV 0.69DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69kev: open, trainable Jev-like family of small decision models on Qwen3.5/3.8A family of small decision models (0.8B to 27B) built on Qwen3.5/3.8 that reproduces Jev's prefill-only typed-decision…models · JEV 0.68GPT-Synopsys: OpenAI and Synopsys multi-year deal to build an AI model that operates EDA toolsOpenAI and Synopsys signed a multi-year agreement on Sep 30 to jointly develop GPT-Synopsys, a specialized model that…models · JEV 0.66
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.