# OpenBMB/MiniCPM

MiniCPM5-1B: A SOTA 1B on-device LLM, small yet powerful.

Repository: https://github.com/OpenBMB/MiniCPM
Canonical: https://ross.abutalabs.com/products/minicpm
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Last push: 2026-07-27T06:50:07+00:00

## Health v2 (maintenance only)
Score: 75/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 94, release rhythm 54, longevity 67
- inputs: {"age_days": 947, "days_push": 37, "days_rel": 99, "gap_med": 328, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 10250, forks 693 (observed 2026-08-28T04:10:41.296073+00:00)

## What it is
MiniCPM is a family of small, state-of-the-art on-device language models from OpenBMB, with MiniCPM5-1B being a dense 1B Transformer for local and resource-constrained deployment. The repository provides model weights, deployment cookbooks, fine-tuning guides, and agent skills for major inference backends.

## Use cases
- run a small LLM locally on-device
- deploy a 1B language model on resource-constrained hardware
- fine-tune a compact open-source LLM
- build an agent with local tool-use model
- hybrid reasoning chatbot that can toggle thinking mode
- efficient inference with million-token context

## When to choose
- you need a powerful small LLM that runs on-device or offline
- you want open-source SOTA quality in the 1B size class
- you need both fast chat and deliberate reasoning from one checkpoint
- you want cookbooks for multiple inference and fine-tuning backends

## When to avoid
- you need frontier-level reasoning only large models provide
- you want a hosted API rather than self-managed weights
- you need multimodal vision capabilities, which live in the separate MiniCPM-V repo

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, llm-training, machine-learning, agent-framework
- domain: large-language-models, artificial-intelligence
- platform: python, cross-platform, self-hosted
- tags: on-device-llm, small-language-model, hybrid-reasoning, model-weights, fine-tuning, edge-deployment, natural-language-processing, gpu

## Member repositories
- OpenBMB/MiniCPM (main) score 75

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:41.296073+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:19:24.445243+00:00, confidence not recorded.
  - readme: https://github.com/OpenBMB/MiniCPM (fetched 2026-08-28T04:10:41.296073+00:00, sha fd752f7dc805)
- Data as of 2026-08-30T08:39:29.467469+00:00.