W
wattai
A small local model that picks from your options and says how sure it is
Emerged finds new products worth a look, daily — what’s new today
If you pay an LLM to pick a department or flag churn risk on every ticket, a 400M-parameter model on a T4 can do it in about 33 ms with no output to parse.
Laya is a Python library and three open checkpoints for structured decisions over text, email or JSON: pick a label, score on a scale, or answer yes or no. Nothing is generated, so every answer comes back typed with a confidence. A router detects the language and sends each request to an English model on ModernBERT-large or a multilingual one on mmBERT-base. Install is pip, with a Colab notebook and a Hugging Face demo. The README is frank about weak spots: ordinal scoring, label sets above fifty options, and near-chance zero-shot results before fine-tuning.
More like this
SemIf
Read a yes/no probability straight out of a small open model, no generated text
Kev
Open 0.8B to 9B decision models for choice, yes/no and rating questions
jev-align
Teach a Jev classifier your labels by reviewing only its least certain answers
L
Laya Vision
A 256M model that answers typed questions about an image with calibrated odds
jevals
Run dozens of agent evals per trace in one sub-second request, no judge prompt