whereismyai
Gutsy
Gutsy: Local decision model that provides calibrated probabilities for yes/no, choice, and score questions based on provided text or JSON state inputs.
Gutsy is a local decision model that calculates probabilities for yes/no, choice, and score questions without performing network calls or generating text. Users send a state in text or JSON format to the system, which returns a probability for every provided option. It runs on ordinary CPUs using llama.cpp and the gutsy-inference runtime, ensuring data stays on the local machine. The model is a fine-tune of Qwen3.5-0.8B, providing deterministic outputs that remain consistent even when option order is shuffled. It handles up to 255 options per request and returns a rejection probability if none of the provided choices fit the state.
Gutsy at a glance
- Best for: Software engineer
- Runs on: CLI, self-hosted
- Pricing: Open source; Apache 2.0 license.
- Released: October 2026
- Category: AI research tool
- Website: github.com/kouhxp/gutsy
Key features
- Local execution on ordinary CPUs via llama.cpp
- Calibrated probability scoring for input states
- Support for up to 255 options per request
- Deterministic output generation
- Support for yes/no, choice, and score question types
- Privacy-focused operation with no network calls
- Rejection probability for when no options fit
Use it for
- Classifying intents for support triage
- Automating model routing for message processing
- Validating claims against provided source data
- Executing triggers during sales calls
Alternatives to Gutsy
- NanoJev — It runs parallel decisions and outputs probability distributions without output-token decoding.
- Decido — Runs locally on CPU while beating published benchmark scores.
- Vegaml — Runs locally on your hardware with calibrated probabilities and conformal answer sets.
- SemIf OpenJev — Extracts direct token probabilities without generating text or running JSON repair loops.
- Clef — Offers low latency and high-speed classification for agentic workflows.
- LLM Stats — Independent rankings updated continuously from live API metrics.
- Laya — It runs typed decisions in a single forward pass over 100+ languages.
- Kev — Provides calibrated probabilities and matches the TypeSafe API.
- PaperLeet — Practice implementing 89 Python exercises derived directly from AI research papers.
- Vev — Runs locally on NVIDIA hardware with a fast, Jev-compatible API.
- HyperResearch — Generates verified reports with citations checked against full source texts and a persistent vault.
- OpenDecision — It uses SigLIP2 to quickly match images against candidate captions and rules.
All Gutsy alternatives