whatmodel.sh / MiniCPM3-4B

MiniCPM3-4B

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

MiniCPM3-4B is an open-weights language model from OpenBMB, released September 2024. Licence: minicpm-model-license.

Catalogue identity notes

Third-generation MiniCPM text model (maker repo openbmb/MiniCPM3-4B @d6b14dd, repo created 2024-09-03). Card front matter says apache-2.0 for the repository code, but 'The usage of MiniCPM3-4B model weights must strictly follow MiniCPM Model License.md' (free academic use; commercial use after registration), hence licence minicpm-model-license as for minicpm-v-2-6. Parameter count not stated by the maker in the card and no safetensors metadata, so params_b is left out. Appears in BFCL v3 (2025-06-14) and V4 (2025-12-16) as 'MiniCPM3-4B-FC (FC)' / 'MiniCPM3-4B (Prompt)', and in the LiveCodeBench website release 2024-12-26 (aa37a9f).

Model information sources

Release date ↗ · Weights licence ↗

47.5±2.1#390 of 719 LLMsfrom 3 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of MiniCPM3-4B, out of 828 with a score, then more from OpenBMB.

#modelscorewhereRAMreleased
465Qwen2.5 Coder 3B Instruct Alibaba48±1local~3.3GB2024-11
466DeepSeek-Coder 6.7B Instruct DeepSeek48±1local~5.8GB2023-10
467MiniCPM3-4B OpenBMB48±2local~4GB2024-09
468Bielik 4.5B v3.0 Instruct SpeakLeash48±2local~4.3GB2025-04
469Qwen3.5 2B Alibaba47±1local~2.9GB2026-03
more from OpenBMB
186MiniCPM5 2B OpenBMB64±1local~3GB2026-09
574MiniCPM-V 2.6 OpenBMB43±2local~6.4GB2024-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #258 of 300 * · 25% of its score · 2 tests

Beats DeepSeek R1 0528 Qwen3 8B, Magistral Small 1.0 (2506), Llama 4 Scout (109B-A17B) here, though they rank higher overall. Trails Granite 4.0 H 1B, Granite 4.0 1B, Qwen2.5 3B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BFCL V4 (December 2025 archive)25.6% ✎▁▁▁▂██▅▅▆▂▂▂ beats 15/77hard · avg 32%▇▇▇▇▇▇ 16%at its level
BFCL v350.3% ✎▁▁▁▁▁▃▅█▅▄▁▁ beats 34/114medium · avg 52%▇▇▇▇ 8.8%blunt · at its level
coding▁▁▁▃▆██▇▆▄▃▁ #287 of 458 * · 73% of its score · 1 test

Beats Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B), GLM-4.7-Flash here, though they rank higher overall. Trails DeepSeek LLM 67B Chat, WizardLM 70B V1.0, Yi 1.5 9B Chat here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LiveCodeBench official (Jan 1–Nov 1, 2024)21.5% ✎▁▁·▁▂▄▆█▅▂·▂ beats 6/31hard · avg 26%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 73%sharp

$ where to run it

llama.cpp · recommended

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run MiniCPM3-4B · needs ~2.47 GB · from huggingface.co/openbmb (official)
    chat in the terminal
    llama cli -hf openbmb/MiniCPM3-4B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf openbmb/MiniCPM3-4B-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run MiniCPM3-4B · needs ~2.47 GB · from huggingface.co/openbmb (official)

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/openbmb/MiniCPM3-4B-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run MiniCPM3-4B · needs ~2.47 GB · from huggingface.co/openbmb (official)
    chat in the terminal
    ollama run hf.co/openbmb/MiniCPM3-4B-GGUF:Q4_K_M

MLX (Mac)

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run MiniCPM3-4B · needs ~2.4 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/MiniCPM3-4B-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/MiniCPM3-4B-4bit

Official weights and documentation: huggingface.co/openbmb ↗

$ how the score was worked out · LiveCodeBench official websit… 73%, Berkeley Function Calling Lea… 25%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LiveCodeBench official website results (2024-12-26)1▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 73%CC-BY-SA-4.0
Berkeley Function Calling Leaderboard result archive2▇▇▇▇▇▇▇ 25%Apache-2.0
prior · pull toward the middle▇ 2%
weighted average= 47.5

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json