MiniCPM5-2B
A 2-billion-parameter small language model released by OpenBMB, trained using OPD (Omni-domain Policy Distillation) and reinforcement learning. It supports a 128K context window, tool calling, and code reasoning. The model is built on a LLaMA-based architecture and is open-sourced under the Apache 2.0 license for commercial use.
Timeline (last 90 days)
MiniCPM5-2B is available for download via ModelScope.
MiniCPM5-2B features a 128K native context, tool calling, code reasoning, mathematical reasoning, and long-text understanding capabilities.
OpenBMB used reinforcement learning combined with OPD (Omni-Policy Distillation) to merge the capabilities of 16 expert models (including 5 agent experts) into MiniCPM5-2B.
MiniCPM5-2B open-sourced its training data, including approximately 500,000 agent samples and over 80,000 reinforcement learning samples.
MiniCPM5-2B adopts the standard LLaMA architecture and can be used out of the box via llama.cpp without an API key or internet connection.
MiniCPM5-2B uses the Apache 2.0 license, and no additional application is required for commercial use.
MiniCPM5-2B's score on SWE-bench beats 4B-class competitors with twice its parameter count.
MiniCPM5-2B is a 2-billion-parameter model launched by OpenBMB.
MiniCPM5-2B achieved a score of 46.4 on SWE-bench.
A 2B-parameter model after INT4 quantization typically requires only about 1–2 GB of memory and can run on mainstream smartphones, NVIDIA Jetson series, and other edge devices.
5 more timeline events
All Facts (14)
A 2B-parameter model after INT4 quantization typically requires only about 1–2 GB of memory and can run on mainstream smartphones, NVIDIA Jetson series, and other edge devices.
60%UnverifiedMiniCPM5-2B is available for download via ModelScope.
50%UnverifiedMiniCPM5-2B open-sourced its training data, including approximately 500,000 agent samples and over 80,000 reinforcement learning samples.
50%UnverifiedMiniCPM5-2B adopts the standard LLaMA architecture and can be used out of the box via llama.cpp without an API key or internet connection.
50%UnverifiedMiniCPM5-2B uses the Apache 2.0 license, and no additional application is required for commercial use.
50%UnverifiedMiniCPM5-2B's score on SWE-bench beats 4B-class competitors with twice its parameter count.
50%UnverifiedMiniCPM5-2B is a 2-billion-parameter model launched by OpenBMB.
50%UnverifiedMiniCPM5-2B achieved a score of 46.4 on SWE-bench.
50%UnverifiedMiniCPM5-2B features a 128K native context, tool calling, code reasoning, mathematical reasoning, and long-text understanding capabilities.
50%UnverifiedOpenBMB used reinforcement learning combined with OPD (Omni-Policy Distillation) to merge the capabilities of 16 expert models (including 5 agent experts) into MiniCPM5-2B.
50%UnverifiedA model with 2B parameters typically requires only about 1–2 GB of memory after INT4 quantization, enabling it to run on mainstream smartphones, embedded development boards, and similar devices.
50%UnverifiedMiniCPM5-2B is a small language model with 2B parameters.
50%UnverifiedThe OpenBMB team released MiniCPM5-2B, a small language model with only 2B parameters.
50%UnverifiedMiniCPM5-2B is fully open-sourced on Hugging Face and GitHub.
50%