BlueLM is the family of open large language models developed by vivo AI Lab, with 943 stars on GitHub and downloadable weights for research and product use.
BlueLM is vivo’s open large language model series, released by the smartphone maker’s AI lab for the wider developer community. The flagship BlueLM-7B models come in base and chat variants, and the chat model was extended to a 32K context window with function calling support, letting it drive tool-using applications rather than just conversation. Quantized AWQ and GPTQ versions are published alongside the full-precision weights so the models can run on more modest hardware. The repository includes inference and deployment code โ including an OpenAI-format API server โ plus fine-tuning support, and the weights are distributed through Hugging Face, ModelScope and WiseModel under the OpenAtom Model License. The project was last updated in late 2024. Evaluation results published in the repository compare BlueLM against other open models of a similar size on standard benchmarks, so teams can judge whether it fits their workload before downloading any weights.
BlueLM is free: the model weights and code can be downloaded and used under the terms of the OpenAtom Model License, with no API fees or subscriptions. Any costs are only the hardware you choose to run or fine-tune the models on.
BlueLM is for developers and researchers who want a capable open Chinese-English language model from an established device maker, especially for on-device or self-hosted scenarios. It also suits teams evaluating function-calling open models who need long-context chat behaviour without paying per token.