Organization repository on GitHub
Jeeves – Reasoning improves Jev-like decision models
- Forks
- 21
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Models & Reimplementations
Jeeves
PostHog's reasoning Jev-style classifier with a diffusion drafter, reporting higher accuracy than Kev-9B and hosted Jev on held-out and JevBench tiers.
- Stars
- 420
- Language
- Python
- License
- MIT
- Published by
- Community
- Author
- PostHog
- Last verified
- Oct 2026
About this project
A reasoning Jev-style classifier with a diffusion drafter, trained with SFT and CISPO. A 9B Jev-like model (Qwen3.5-9B, LoRA, pointer head) that thinks before it decides, with a block-4 diffusion drafter and the full training code and train/dev/test data. Beats Kev-9B and Jev on test data it was never trained on (0.889 vs 0.822 and 0.857) and on JevBench's public tiers (0.935 vs 0.866 for Jev).
Quoted from the project's README. Read the full README.
- Languages
- Python
- Capabilities
- Open source
Highlights
- README cites 0.889 test accuracy versus Kev-9B 0.822 and Jev 0.857 on data outside training.
- JevBench public tiers reported at 0.935 versus 0.866 for hosted Jev on hard.
- Exposes noul, choice, and score questions through a Jev-compatible API with a thinking cap.
Watch out
MIT research codebase with training data bundled; numbers are author-reported, not an independent audit.
Questions about this entry
Short answers drawn from the linked project or page.
Both, according to the README. The same checkpoint scores 0.804 without thinking and 0.840 with it on a 2,962-item test split, at about 0.3 s per request without thinking versus a 3.3 s median with it on one H100 in fp8. Truncating the thinking chain is the documented way to buy the latency back.
Only some rows. Transfer overall, which combines MMLU-Pro and buried state, is 0.746 against 0.800 for Jev, and plain MMLU is 0.793 against 0.900. The wins are on the author's own test split (0.889 against 0.857) and on JevBench's 231 public items (0.935 against 0.866).
Only the public easy, standard and hard tiers; the sealed judge tier is not included. The comparison columns for the other model are numbers that project publishes, and the asterisked rows are its 8B Qwen3 checkpoint rather than the 9B one, because no 9B JevBench result has been published. All of it is author-reported.
On a Mac, yes, but not comfortably at defaults: the weights take 21 GB in bf16 and the default caches add another 28 GB, so a 48 GB machine starts to swap until you lower the row and length limits. Precision fp8 drops weights to about 11.5 GB and speeds up thinking, at the cost of slightly different outputs.
A 9B Jev-like model, Qwen3.5-9B with LoRA and a pointer head, trained with SFT and CISPO and paired with a block-4 diffusion drafter. It is MIT-licensed and supports noul, choice, and score questions in one request through a Jev-compatible API.
Download with hf download PostHog/jeeves --local-dir jeeves-weights, then run python -m inference.serve --model jeeves-weights --drafter jeeves-weights/drafter_k4.safetensors --port 8009. Requests go to POST localhost:8009/v1/systemone in Jev's format.
The sdk/ directory is a drop-in replacement for Jev's Python SDK (typesafe-sdk), installed with pip install ./sdk and exposing choice, noul, and score questions plus return_reasoning. You can also use curl or any client that speaks the System One request shape.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
Hacker News: Decision Index 0.2.1 scores open reproductions of Jev on a frozen suite.
jackbravo4 pointsDecision Index 0.2.1 – 70 open reproductions of TypeSafe Jev's Decision ModelRead the thread on Hacker NewsHacker News: Pointer to a DiffusionGemma plus vLLM tutorial for a Jev-shaped classifier.
1 pointHow to build a Jev-style classifier with DiffusionGemma and vLLMRead the thread on Hacker NewsHacker News: A one-token logprob wrapper that also classifies webcam frames.
allanrbo151 pointsA single function Jev-like wrapper for LLMs, including vision modelsRead the thread on Hacker NewsRelated terms
Glossary definitions related to this entry.
More Jev resources like this
59GitHub stars
A preregistered measurement of a Jev-style interface on frozen Qwen3-4B: read the option letter's logits instead of generating grammar-constrained JSON.
Jev Models & Reimplementations#community#python#local
Community
r-ms/mini-jevPython
r-ms/mini-jevPythonThe fine-tuned Laya checkpoint for four typed-decision workflows: 0.766 accuracy on 2,000 held-out decisions, above Jev 1.13.0's published 0.727 and the teacher self-agreement ceiling it trained past.
Jev Models & Reimplementations#community#open-models#local
Community
877GitHub stars
A local server that reads next-token probabilities from a fine-tuned Spark model and returns Choice, Score, and Noul answers with zero generated tokens.
Jev Models & Reimplementations#community#python#open-models
Community
Rizzo-AI-Academy/rizzo-flowPython
Rizzo-AI-Academy/rizzo-flowPython130GitHub stars
An open recipe for a 27B multimodal decision model trained with full-weight SFT on 73,000 examples over one H200, serving choice, noul and score on /v1/systemone with per-checkpoint provenance and released calibration plots.
Jev Models & Reimplementations#community#calibration#qwen
Community
denis-pplx/autojevPython
denis-pplx/autojevPython58GitHub stars
A local /v1/systemone server that reads label probabilities from a GGUF chat model, with prompt repetition and a confidence shrink toward uniform.
Jev Models & Reimplementations#community#python#open-models
Community
Mushroom-Systems/lichenPython
Mushroom-Systems/lichenPython10GitHub stars
A local Jev-compatible server that reads Choice, Score and Noul answers from the logits of any MLX checkpoint or OpenAI-compatible endpoint with no training, serving POST /v1/systemone as a drop-in for the official TypeSafe SDK.
Jev Models & Reimplementations#community#mlx#calibration
Community
bradAGI/rulingPython
bradAGI/rulingPythonFrom the Jev guides
Original write-ups on what this entry is for.
Interface, size, license, claims — in that order. A selection guide across the system-one catalog, from a 706K form scorer to 27B Jev students.
Jev Practices & Patterns#open-models#evaluation#benchmarks
Read article
Base decision checkpoints sit near chance on purpose. The typed-decisions loop, the corpora behind the students, and the temperature fit that turns probabilities into thresholds.
Jev Practices & Patterns#fine-tuning#calibration#open-models
Read article
Hundreds of repositories carry the system-one tag, but the ecosystem is a dozen ideas in different sizes. A tour of the families, the licenses, and what a download does not include.
Jev Models & Reimplementations#open-models#hugging-face#local
Read article
From the Jev community
Posts from builders shipping with Jev right now.