Quyet-1.2-Small
Quyet-1.2-Small is a decision model: given a state (any text, JSON or conversation) and one or more typed questions, it returns temperature-scaled probabilities over the supplied options. Question types: choice (pick one option), score (an ordered scale) and noul (true / false). It is part of the Quyet 1.2 family (Large, Small, Tiny), released by Chinh Nguyen under Apache-2.0.
| Base model | google/gemma-4-E4B-it |
| Architecture | Full fine-tune of the Gemma 4 E4B text decoder + scalar option head |
| Parameters | 7.46B (7,463,015,937) |
| Languages | Multilingual |
| Input | 23,434 complete tokens; no truncation |
| Weights | 14.9 GB (BF16), self-contained |
| License | Apache-2.0 (see LICENSE and NOTICE) |
| Homepage | quyet.ai |
New in 1.2
- A new model, built on the Gemma 4 E4B text model.
- Up to 256 options per choice question; 1.0 allowed 10.
- Inputs up to 23,434 tokens, never truncated; an input over the limit is rejected instead of being cut.
- Opt-in shared prefix caching, useful when several questions share a long state.
- The model's files are verified when it loads, and every weight hash can be checked on request.
How to use
pip install "quyet[student]==1.2.0"
import quyet
m = quyet.load("chinhnc/Quyet-1.2-Small") # downloads from Hugging Face
r = m.predict(
{"message": "Please close my card, I lost it yesterday."},
{"intent": {"type": "choice", "instructions": "What does the customer want?",
"criteria": {"cancel": "close the card", "limit": "change the limit", "other": None}},
"urgent": {"type": "noul", "instructions": "The request is urgent."},
"mood": {"type": "score", "instructions": "How upset is the customer?", "criteria": ["calm", "annoyed", "angry"]}},
)
print(r["answers"]) # {"intent": {"choice": ..., "confidence": ..., "probabilities": {...}}, "urgent": {"noul": P(true)}, ...}
The model folder contains all text weights and its tokenizer; no separate base download is needed. Runs on one GPU in BF16 or on CPU in float32; device_map is not supported. On NVIDIA H100 80GB HBM3 in BF16, a check used about 16.5 GiB of allocated GPU memory and 26 GiB reserved; a short float32 CPU check used about 34 GiB of process memory. Memory use depends on the input and batching. Enable the opt-in cache with prefix_cache=True when the shared state is long; it requires transformers 5.17.0. Choice questions use the scalar option head; score and noul questions use vocabulary-symbol scores. The package builds the input and applies the configured temperature.
Answers contain choice (with confidence and probabilities), score (expected level, confidence, probabilities, legend) or noul (P(true), with confidence). Each choice question accepts 2 to 256 options, score 2 to 10 levels, and noul exactly 2 options. Inputs are never truncated; inputs exceeding the complete-token limit are rejected.
Not evaluated for safety, bias or adversarial robustness; do not use as the only control for high-stakes decisions. In half precision, results can differ slightly between hardware and batch sizes on near-tie questions.
Credits
- Gemma 4 by Google (Apache-2.0): https://huggingface.co/google/gemma-4-E4B-it.
Citation
@misc{quyet2026,
title = {Quyet 1.2: calibrated decision models},
author = {Chinh Nguyen},
year = {2026},
url = {https://huggingface.co/chinhnc/Quyet-1.2-Small}
}
License
Apache-2.0. Keep the NOTICE file (it starts with "Quyet by Chinh Nguyen") when you redistribute this model or anything derived from it. Questions and issues: email@quyet.ai. Package source: ncchinh/quyet.
- Downloads last month
- 37