Winnow-12B
Open-weight GGUF decision model that reads answer probabilities from its logits in one pass. A Jev-class model you can run locally.
0 members list Winnow-12B, unchanged over the last 12 weeks, data as of 9 November 2026
Members who list Winnow-12B in their Stack, by week.
Members who list Winnow-12B
Be the first to add Winnow-12B
Nobody on Stackness runs Winnow-12B yet. Add it to your stack and you're the first member running it.
About Winnow-12B
Winnow-12B is an open-weight language model available in GGUF format that performs decision tasks by extracting answer probabilities from model logits in a single inference pass. The model supports image and text inputs, enabling developers to run typed decision-making, chat, and vision capabilities locally on their machines. It integrates with popular local inference frameworks including llama.cpp, vLLM, Ollama, and other tools, making it accessible for developers who want to deploy AI without relying on external services.
The model was developed by EldanRing and is licensed under the MIT license. It is written in Python and designed to work with standard CPU and GPU hardware setups, supporting configurations that allow for large context windows with vision capabilities. As an open-source project, Winnow-12B enables developers to experiment with and customize the model for their specific use cases.
Checked 2 November 2026.
Nobody on Stackness runs Winnow-12B yet.
Sign up and add itPosts about Winnow-12B
- How do you run System One decision models locally? Ollaya, laya-mlx and the runtime slot
To run a System One model locally, pick a runtime too: Ollaya serves ten open model families, laya-mlx runs Laya on Apple silicon, and llama.cpp needs a shim.
- How do you benchmark a System One model? What JevBench scores, and what calibration says it cannot
JevBench scores System One models on accuracy, calibration, speed and cost, and puts Jev first at 63.3. Why calibration still needs your own data.