On February 18th, Enflame, a domestic AI chip company, announced the completion of deep adaptation work for several mainstream domestic open-source large models, including Zhipu's GLM-5.0 and Alibaba's Qwen3.5-397B-A17B. To date, Enflame has successfully adapted over 40 AI large models for immediate deployment, including DeepSeek, Qwen, GLM, Intern-S1, and Wenxin. These models cover various domains such as large language models, vector models, multimodal understanding models, multimodal generative models, and mainstream OCR models. All adaptation work is conducted on its self-developed T100 acceleration card, utilizing the tiered development toolchain within the SDAA software stack. This approach comprehensively addresses diverse development needs from beginner to advanced levels, enabling developers to quickly build high-performance operators, achieve seamless compatibility with mainstream AI ecosystems, and significantly reduce the technical barriers and costs associated with migrating from the CUDA ecosystem.