Show HN: Gutsy – small decision model for CPU (500MB), get probabilities back

Category: ai-ml

Tags: decision-model, probability-calibration, local-inference, cpu-inference, llama-cpp, gguf

Score: 7.5/10 (Innovation: 7, Technical: 8, Documentation: 7, Utility: 8)

Gutsy is a small, CPU-friendly decision model (fine-tuned Qwen3.5-0.8B) run via llama.cpp that returns calibrated probabilities for typed questions (yes/no, choice, score) given a state. It emphasizes local inference, determinism, option-order robustness, and state caching, making it practical for low-latency decision tasks without text generation. Its Jev-style API eases adoption.

Target audience: backend developers, ML engineers, and data scientists needing local probabilistic decision support

Repository: https://huggingface.co/kouhxp/gutsy · C++ · MIT · 131948 stars

View on Hacker News