Jeeves
Jeeves is a 9B reasoning classifier based on Qwen3.5-9B, with a pointer head and a block-4 diffusion drafter. Released weights, training code, and a Jev-compatible API are provided.
What it does
Use it to produce calibrated decision probabilities and classifications, optionally reasoning before deciding when confidence is low.
- Availability
- Weights downloadable from Hugging Face; inference instructions and code provided
- Pricing model
- Not established
- Deployment
- Runs on CUDA GPUs and Apple Silicon using MPS
- Privacy
- Not established
- Licence
- MIT
Capability
Modality
Access
Deployment
Pricing
Target user
Evidence
6 claims- Jeeves supports yes/no, multiple-choice, and rating questions in the same request. [1]
- Supports yes/no (`noul`), multiple-choice (`choice`), and rating (`score`) questions in the same request, through a Jev-compatible API.
GitHub new repositories. Quoted text is outside CC BY 4.0 and keeps its original rights. - Jeeves can answer directly when confidence is high and reason first when confidence is low. [1]
Jeeves v2 and v1 answer directly when their calibrated confidence is at least 0.9 and otherwise reason first (`nothink_threshold` 0.9, greedy, 2,560-token cap); the percentage is the share of questions that reasoned.
GitHub new repositories. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model weights are available from PostHog on Hugging Face. [1]
- Inference runs on CUDA and Apple Silicon using MPS. [1]
Inference also runs on Apple Silicon (MPS) in bf16 or FP8.
GitHub new repositories. Quoted text is outside CC BY 4.0 and keeps its original rights. - The project provides a Jev-compatible API and a drop-in replacement for Jev's Python SDK. [1]
`sdk/` is a drop-in replacement for Jev's Python SDK (`typesafe-sdk`):
GitHub new repositories. Quoted text is outside CC BY 4.0 and keeps its original rights. - The repository lists the model's licence as MIT. [1]
Sources
1
1
GitHub new repositoriesOwner-controlled
https://github.com/PostHog/jeeves
Retrieved 2026-10-28 00:02 UTC
Verified
Versions
4| Version | Effective | Change | Changed fields | State |
|---|---|---|---|---|
| 4 | 2026-10-17 03:52 UTC | material change | capability, availability, deployment | Current |
| 3 | 2026-10-01 18:33 UTC | material change | access, pricing, availability, deployment | Current |
| 2 | 2026-09-30 19:27 UTC | material change | capability, access, deployment, availability, deployment | Current |
| 1 | 2026-09-29 20:20 UTC | publication | None | Current |
In briefs
1- OpenAI introduces GPT-6.1 Sol for coding and multi-step workflows Daily brief 2026-09-29
Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.