MiniCPM5-1B is now live — the strongest open‑source base model under 2B.🚀🚀
🔥 Ranks #1 on the Artificial Analysis (AA) index for small models, scoring 17.9 to beat the 2B-scale Qwen3.5-2B (16.3).
⚡ Comprehensively surpasses Qwen3.5-0.8B and LFM2.5-1.2B-Thinking in knowledge, math, coding, and tool use.
🏗️ INT4 weights = just ~0.5GB — runs on phones, browsers, laptops.
👾 Powers fully offline AI “Desktop Pet” — no cloud, no GPU cluster.
Try the model here:
🤗 Hugging Face: huggingface.openbmb.com/model/openbmb/…
💻 GitHub: github.com/OpenBMB/MiniCPM
🔭 Modelscope: modelscope.cn/models/OpenBMB…
🔥 Ranks #1 on the Artificial Analysis (AA) index for small models, scoring 17.9 to beat the 2B-scale Qwen3.5-2B (16.3).
⚡ Comprehensively surpasses Qwen3.5-0.8B and LFM2.5-1.2B-Thinking in knowledge, math, coding, and tool use.
🏗️ INT4 weights = just ~0.5GB — runs on phones, browsers, laptops.
👾 Powers fully offline AI “Desktop Pet” — no cloud, no GPU cluster.
Try the model here:
🤗 Hugging Face: huggingface.openbmb.com/model/openbmb/…
💻 GitHub: github.com/OpenBMB/MiniCPM
🔭 Modelscope: modelscope.cn/models/OpenBMB…
🧸 1B scale, real‑world execution.
MiniCPM5‑1B can run a complete, offline AI Desktop Pet entirely on your local device.
No internet? No problem. No GPU? Still works.
This is what “small model, big potential” really means.
MiniCPM5‑1B can run a complete, offline AI Desktop Pet entirely on your local device.
No internet? No problem. No GPU? Still works.
This is what “small model, big potential” really means.
High Intelligence Density 📈
🔹 Performance: MiniCPM5-1B secures #1 on the Artificial Analysis (AA) small models index (17.9). It is the strongest open-source base model under 2B, and the strongest non-reasoning model under 4.5B.
🔹 Efficiency: Reaches a 17.9 AA score using only ~15M output tokens. To achieve a lower score of 16.3, Qwen3.5-2B consumes over 268M tokens. That is a 17x reduction in token budget for superior intelligence.
🔹 Performance: MiniCPM5-1B secures #1 on the Artificial Analysis (AA) small models index (17.9). It is the strongest open-source base model under 2B, and the strongest non-reasoning model under 4.5B.
🔹 Efficiency: Reaches a 17.9 AA score using only ~15M output tokens. To achieve a lower score of 16.3, Qwen3.5-2B consumes over 268M tokens. That is a 17x reduction in token budget for superior intelligence.
AI Forging AI 🛠️
The base model was pre-trained using ForgeTrain, the world’s first fully AI‑generated production‑level pre‑training framework.
No human in the loop, ForgeTrain achieved training speeds 10% faster than Nvidia's Megatron.
The base model was pre-trained using ForgeTrain, the world’s first fully AI‑generated production‑level pre‑training framework.
No human in the loop, ForgeTrain achieved training speeds 10% faster than Nvidia's Megatron.
Built for Developers
🔹 Deployment: In INT4 quantization, the model weight is just ~0.5GB. It runs natively on phones, tablets, and even directly inside a web browser with zero configuration.
🔹 Fine-Tuning Support: LLaMA-Factory, ms_swift.
🔹Inference Frameworks: Out-of-the-box compatibility with SGLang, vLLM, llama.cpp, Ollama, Hugging Face, and ArcLight
#MiniCPM #EdgeAI #OnDeviceAI #OpenSourceAI #LLM #SmallModels #AI #LocalLLM #GenerativeAI #OpenBMB
🔹 Deployment: In INT4 quantization, the model weight is just ~0.5GB. It runs natively on phones, tablets, and even directly inside a web browser with zero configuration.
🔹 Fine-Tuning Support: LLaMA-Factory, ms_swift.
🔹Inference Frameworks: Out-of-the-box compatibility with SGLang, vLLM, llama.cpp, Ollama, Hugging Face, and ArcLight
#MiniCPM #EdgeAI #OnDeviceAI #OpenSourceAI #LLM #SmallModels #AI #LocalLLM #GenerativeAI #OpenBMB
• • •
Missing some Tweet in this thread? You can try to
force a refresh
