GitHub radar
GLiNER2.5-Decide: a tiny model that sorts your messages
Fastino released a 340M classifier that takes any label set you give it at call time — intent, routing, sentiment, priority — and sorts a piece of text in one pass, no fine-tuning per task.
Fastino’s GLiNER2.5-Decide is a 340M parameter multi-label classifier. You give it a list of labels (e.g. customer intents, ticket categories, sentiment, priority, whether it’s spam or not) at runtime, and it will predict multiple labels simultaneously for a given input, without needing prompts or task-specific finetuning. In their internal tests on a diverse set of 17 domains with 300 test examples per domain, they achieve an exact match accuracy of 60.2%, outperforming the much larger Qwen3.5-4B which achieves 56.4% (roughly twelve times smaller). They note that this is not a general purpose model, nor is it capable of reasoning or explaining itself; it simply provides the labels you request. This model is available for CPU/GPU inference via the tiny gliner2 python package.
Why a vibe-coder should care
This kind of model is perfect if you need to classify text data (such as support tickets, reviews, leads) into various categories. Rather than having to pay someone to make calls to large models, you could have an agent write a short script to automate this process for you. Of course, this model is very specialized and doesn’t do anything else besides classifying text into your chosen labels — there’s no generation or explanation here, just classification. But if all you care about is classifying text, then this model is a great choice since it performs better than much larger general purpose models on their internal benchmarks.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the GLiNER2.5-Decide model from Fastino for me: https://huggingface.co/fastino/GLiNER2.5-Decide — install the gliner2 package, write a small script that sorts my incoming messages into categories like complaint, question and spam, and show me the result on three examples.
A regular laptop with no GPU is enough — the model weighs about 1 GB and, per the authors, runs on either CPU or GPU.
Open on Hugging Face▌ More finds