GitHub radar
MiniCPM5-2B: A Small Local Model for Everyday Tasks
OpenBMB released MiniCPM5-2B, a small model built for on-device use that runs through Ollama or LM Studio — I review which routine tasks it can take off your API bill.
MiniCPM5-2B is the second model in OpenBMB's MiniCPM5 series, a 2-billion-parameter model built for on-device and local use. OpenBMB says it leads open models of its size in their comparison and stays competitive with 4B-class models, with strengths in coding, maths, long-context understanding, tool use and agentic tasks. This repository holds the GGUF files for llama.cpp, Ollama and LM Studio: the compressed Q4_K_M version is about 1.6 GB, the 8-bit one about 2.7 GB. OpenBMB also released the training datasets behind the model, and the weights are under the Apache 2.0 license.
Why a vibe-coder should care
Small local models are where I look first when a token bill starts to grow. Sorting emails, drafting replies, pulling fields out of documents — a lot of that goes to a paid API by habit, while a 2B model on a laptop can do it for free and without sending customer data anywhere. The Apache 2.0 license also lets you build it into a product. A good first project: a local assistant that pre-sorts incoming requests before anything reaches the expensive model.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up MiniCPM5-2B locally from https://huggingface.co/openbmb/MiniCPM5-2B-GGUF — use the Q4_K_M version, run it with Ollama and show me how to send it simple tasks instead of a paid API.
Обычный ноутбук от 8 ГБ оперативки — сжатая версия модели весит около 1,6 ГБ.
Open on Hugging Face▌ More finds