Downloads · 30 days
97
31% of all-time downloads
MiniCPM5-2B-MLX-BF16
junwatu/MiniCPM5-2B-MLX-BF16
MiniCPM5-2B-MLX-BF16 is a text generation model from junwatu. Use it when you need the model to write or continue text. It is set up for mlx. The card lists the license as apache-2.0.
MLX BF16 conversion of openbmb/MiniCPM5-2B for Apple Silicon. No quantization — pure BF16. For the official 4-bit MLX release, see openbmb/MiniCPM5-2B-MLX.
All-time downloads
308
Public
Parameters
2.5B
5 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors5 GB · 100%
At a glance
- Task
- Text Generation
- Library
- mlx
- License
- apache-2.0
- Model type
- llama
- Access
- Public
- Created
- Sep 8, 2026
- Updated
- Sep 8, 2026
- SHA
- 724f3ba8
Try a prompt
Base models
Model card
From the Hugging Face model README
MiniCPM5-2B-MLX-BF16
MLX BF16 conversion of openbmb/MiniCPM5-2B for Apple Silicon. No quantization — pure BF16. For the official 4-bit MLX release, see openbmb/MiniCPM5-2B-MLX.
- Architecture:
LlamaForCausalLM, 2.5B params, 42 layers, GQA 16Q/2KV, 131k context - License: Apache-2.0 (same as upstream)
Converted locally with:
mlx_lm.convert --hf-path openbmb/MiniCPM5-2B \
--mlx-path MiniCPM5-2B-MLX-BF16 --dtype bfloat16
Usage
pip install mlx-lm
mlx_lm.generate --model junwatu/MiniCPM5-2B-MLX-BF16 \
--prompt "Who are you?" --max-tokens 128
SGLang (Metal backend):
SGLANG_USE_MLX=1 python -m sglang.launch_server \
--model-path junwatu/MiniCPM5-2B-MLX-BF16 \
--disable-cuda-graph --host 127.0.0.1 --port 30000
- Task
- Text Generation
- Library
- mlx
- Type
- llama
- License
- apache-2.0
- Languages
- en, zh
- Created
- Sep 8, 2026
- Updated
- Sep 8, 2026