MiniCPM5-2B

MiniCPM5-2B

Generates text with a dense 2B-class model made for on-device and local use, covering coding, math, long context, tool use and agent tasks in English and Chinese. Suits developers with limited hardware.

Made by openbmb on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: MiniCPM5-2B is a dense Transformer language model from OpenBMB with about 2.5B parameters by Hugging Face's count. It is the second model of the MiniCPM5 series, after the 1B one, and scales up the same training recipe. It is built for on-device, local and resource-limited use, in English and Chinese.

Who it suits: Developers who want a small model that runs on their own hardware for coding, math, long documents, tool use and agent tasks.

Industry | Who uses it | How they use it | Example
Software | Developers | Run a small coding and agent model locally | Add an offline assistant to a desktop tool
Mobile and edge | Device builders | Put a language model on constrained hardware | Run a chat feature without a server
Education | Students | Try math and coding questions on a laptop | Ask for a worked math solution

If you want the strongest answers and have a large GPU, you can skip it, since the card compares it to 4B-class models.

Where to use it:

• Hugging Face Transformers in Python
• Local and on-device setups, which the model targets
• Windows, macOS or Linux, for downloading the files

Use case | Who | What they get back
Coding help | Developers | Code and fixes from a small model
Math | Students | Worked solutions to math problems
Tool and agent tasks | Agent builders | Tool-using multi-step behavior

How to set it up:

• Python with pip
• Hardware for about 5 GB of BF16 weights, or about 1.3 GB at 4-bit by Hugging Face's estimate
• The Transformers library

1. Install the Hugging Face hub package.
2. Download the model files.
3. Load them with Transformers, following the model card.

pip install -U huggingface_hub
hf download openbmb/MiniCPM5-2B

How to use it: Load the model in Transformers and send it a chat prompt. The card's usage section holds the exact code, and there is an online demo from the same team.

Example: You ask it to write a short function and explain it. The model replies in text with code and a brief note.

For newcomers and experts: If you are new, try one short coding or math question and see how the reply reads. The small size keeps downloads and memory needs low.

The card says the model matches strong open models of similar size in its own comparison set. Across that set, the authors say it keeps pace with larger 4B-class models, with its biggest strengths in code, math, reading long inputs, calling tools and agent work. Those claims come from the authors' comparison, so check them on your own tasks.

Download the weights
Load them in Transformers
Send a chat prompt
Read the generated reply

Good to know:

• It is a dense model, not a mixture of experts
• It supports English and Chinese
• The comparison results are the authors' own
• License: Apache 2.0
• Cost to run: none named, you supply the hardware
• Data it reaches: your prompts and the files downloaded from Hugging Face
• Not for: the card names no banned uses

On Hugging Face, 13 November 2026

• Task: text generation
• Library: transformers
• Parameters: 2.5B
• Weights: about 5 GB in BF16; about 1.3 GB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 1,039,518
• Likes: 1,850
• License: apache-2.0
• Languages: en, zh
• Quantized versions: 131
• Fine-tunes: 90
• Tags: transformers, safetensors, llama, text-generation, minicpm, minicpm5, long-context, tool-calling, on-device, edge-ai, conversational

Links

• Model: https://huggingface.co/openbmb/MiniCPM5-2B
• Files: https://huggingface.co/openbmb/MiniCPM5-2B/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download openbmb/MiniCPM5-2B

Then follow the usage in its model card. Nothing in this kit runs by itself.