🎉 MiniCPM5-1B tops Hugging Face Trending!
Thanks to developers around the world, MiniCPM5-1B — the new on-device text foundation model released by ModelBest together with Tsinghua University and the OpenBMB open-source community — has officially reached No. 1 on the Hugging Face Trending chart! 🏆
💡 Core strengths, three hard-hitting points
🏆 Open-source SOTA at the 1-billion-parameter scale. Compared with strong open-source models of the same size, MiniCPM5-1B reaches SOTA in this comparison set. Its advantages are especially pronounced in agent tool calling, code generation, and complex reasoning tasks.
🧠 Dual-mode chat, smooth even without a GPU. The same core natively supports switching between Think / NoThink modes! Performance soars on math, code reasoning tasks, and tool-calling dimensions.
🔌 Just 0.5GB overhead, 128K ultra-long context. After INT4 quantization, the weights are only 0.5 GB, yet it has a native 128K ultra-long context window! It doesn't hog compute or memory, letting pure CPU devices, phones, and even browsers run a truly local, private large model smoothly.
🐱 An intelligent "desktop pet" for everyone
Runs locally without internet, with low overhead and zero latency, providing an always-available, super-smart desktop companion without sacrificing privacy or system memory.
Every bit of recognition comes from the shared nurturing of the open-source community. The model weights, training datasets, and full deployment suite for MiniCPM5-1B are now fully open-sourced! You are welcome to download and try it on Hugging Face, and we'd love it if you could light up a little star on GitHub 🌟! Your Stars are our greatest motivation to keep pushing the limits of on-device AI!
🔗 Project links:
Hugging Face:
https://huggingface.openbmb.com/model/openbmb/MiniCPM5-1B
GitHub:
https://github.com/OpenBMB/MiniCPM