--- title: MiniCPM5-2B canonical_url: "https://www.modelscope.cn/models/Qualcomm-AI-Hub/MiniCPM5-2B" md_url: "https://www.modelscope.cn/models/Qualcomm-AI-Hub/MiniCPM5-2B.md" repository: Qualcomm-AI-Hub/MiniCPM5-2B last_updated: 2026-10-16 license: apache-2.0 pipeline_tag: text-generation tasks: - text-generation downloads: 0 stars: 0 tags: - llm - generative_ai - android --- # MiniCPM5-2B > MiniCPM5-2B - Qualcomm-AI-Hub 在 ModelScope 开源的模型。MiniCPM5-2B: Optimized for Qualcomm Devices Qualcomm-AI-Hub/MiniCPM5-2B 是 ModelScope 魔搭社区上的text-generation模型，采用 apache-2.0 许可。 - **Repository**: Qualcomm-AI-Hub/MiniCPM5-2B - **License**: apache-2.0 - **Tasks**: text-generation - **Tags**: llm, generative_ai, android - **Downloads**: 0 - **Stars**: 0 - **Last updated**: 2026-10-16 Source: https://www.modelscope.cn/models/Qualcomm-AI-Hub/MiniCPM5-2B --- ![](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/minicpm5_2b/web-assets/model_demo.png) # MiniCPM5-2B: Optimized for Qualcomm Devices For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to https://huggingface.co/openbmb/MiniCPM5-2B This is based on the implementation of MiniCPM5-2B found [here](https://github.com/openbmb/minicpm). This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models-cn/blob/main/src/qai_hub_models/models/minicpm5_2b) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary). Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.cn) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device. ## Deploying MiniCPM5-2B on-device Please follow the [LLM on-device deployment](https://github.com/qualcomm/ai-hub-apps/tree/main/tutorials/llm_on_genie) tutorial. ## Getting Started Below are pre-exported model assets ready for deployment. | Runtime | Precision | Chipset | SDK Versions | Download | |---|---|---|---|---| | GENIE | w4a16 | Snapdragon® 8 Elite Gen 5 Mobile | QAIRT 2.45 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/minicpm5_2b/releases/v0.62.1/minicpm5_2b-genie-w4a16-qualcomm_snapdragon_8_elite_gen5.zip) | GENIE | w4a16 | Snapdragon® 8 Elite Mobile | QAIRT 2.45 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/minicpm5_2b/releases/v0.62.1/minicpm5_2b-genie-w4a16-qualcomm_snapdragon_8_elite.zip) | GENIE | w4a16 | Snapdragon® X Elite | QAIRT 2.45 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/minicpm5_2b/releases/v0.62.1/minicpm5_2b-genie-w4a16-qualcomm_snapdragon_x_elite.zip) | GENIE | w4a16 | Qualcomm® Dragonwing™ IQ-9075 | QAIRT 2.45 | [Download](https://china-qaihub-public-assets.s3.cn-northwest-1.amazonaws.com.cn/qai-hub-models/models/minicpm5_2b/releases/v0.62.1/minicpm5_2b-genie-w4a16-qualcomm_qcs9075.zip) For more device-specific assets and performance metrics, visit **[MiniCPM5-2B on Qualcomm® AI Hub](https://aihub.qualcomm.cn/models/minicpm5_2b)**. ## Model Details **Model Type:** Model_use_case.text_generation **Model Stats:** - Information about the model parts: Prompt Processor and Token Generator are split into 5 parts each. Each corresponding Prompt Processor and Token Generator part share weights. - Input sequence length for Prompt Processor: 128 - Maximum Context length: 2048 - Minimum QNN SDK version required: 2.45 - Num of key-value heads: 2 - Precision: w4a16 is supported - Prompt processor input (other parts): 128 tokens + KVCache initialized with pad token - Prompt processor input (part1): 128 tokens - Prompt processor output (other parts): 128 output tokens + KVCache for token generator - Prompt processor output (part1): Embeddings output - Response Rate: Rate of response generation after the first response token. - Supported languages: Chinese, English. - TTFT: Time To First Token is the time it takes to generate the first response token. This is expressed as a range because it varies based on the length of the prompt. The lower bound is for a short prompt (up to 128 tokens, i.e., one iteration of the prompt processor) and the upper bound is for a prompt using the full context length (2048 tokens). - Token generator input (other parts): 1 input token + past KVCache - Token generator input (part1): 128 tokens - Token generator output (other parts): 1 output token + KVCache for next iteration - Token generator output (part1): Embeddings output - Use: Initiate conversation with prompt-processor and then token generator for subsequent iterations. ## Performance Summary | Model | Runtime | Precision | Chipset | Context Length | Response Rate (tokens per second) | Time To First Token (range, seconds) |---|---|---|---|---|---|--- | MiniCPM5-2B | GENIE | w4a16 | Snapdragon® 8 Elite Gen 5 Mobile | 2048 | 54.79866 | 0.060475 - 0.9676 | MiniCPM5-2B | GENIE | w4a16 | Snapdragon® 8 Elite Mobile | 2048 | 43.12257 | 0.07938733333333334 - 1.2701973333333334 | MiniCPM5-2B | GENIE | w4a16 | Snapdragon® X Elite | 2048 | 7.781388521194458 | 0.09876599999999999 - 1.5802559999999999 | MiniCPM5-2B | GENIE | w4a16 | Qualcomm® Dragonwing™ IQ-9075 | 2048 | 28.8441743850708 | 0.063116 - 1.009856 ## License * The license for the original implementation of MiniCPM5-2B can be found [here](https://github.com/OpenBMB/MiniCPM/blob/main/LICENSE). ## References * [MiniCPM4: Ultra-Efficient LLMs on End Devices](https://arxiv.org/abs/2506.07900) * [Source Model Implementation](https://github.com/openbmb/minicpm) ## Community * For questions or feedback, chat with our engineers in [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.cn). ## Usage and Limitations This model may not be used for or in connection with any of the following applications: - Accessing essential private and public services and benefits; - Administration of justice and democratic processes; - Assessing or recognizing the emotional state of a person; - Biometric and biometrics-based systems, including categorization of persons based on sensitive characteristics; - Education and vocational training; - Employment and workers management; - Exploitation of the vulnerabilities of persons resulting in harmful behavior; - General purpose social scoring; - Law enforcement; - Management and operation of critical infrastructure; - Migration, asylum and border control management; - Predictive policing; - Real-time remote biometric identification in public spaces; - Recommender systems of social media platforms; - Scraping of facial images (from the internet or otherwise); and/or - Subliminal manipulation