Dataphin offers a diverse selection of built-in large models, including those from Alibaba Cloud Model Studio and DeepSeek, to support a wide range of business requirements.
Some models support the deep thinking mode. If the deep thinking mode cannot be disabled for a model, this is noted in the table.
Model service provider | Model display name | Model ID | Description | Is deep thinking supported? |
Alibaba Cloud (Model Studio) | Qwen-Max | qwen-max | The best-performing model in the Qwen series. It is suitable for complex, multi-step tasks. | No |
Qwen-Max-Latest | qwen-max-latest | The best-performing model in the Qwen series. It is suitable for complex, multi-step tasks. Its capabilities are always the same as the latest snapshot version. | No | |
Qwen-Plus | qwen-plus | A balanced model. Its inference performance, cost, and speed are between those of Qwen-Max and Qwen-Turbo. It is suitable for moderately complex tasks. | Yes | |
Qwen-Plus-Latest | qwen-plus-latest | A balanced model. Its inference performance, cost, and speed are between those of Qwen-Max and Qwen-Turbo. It is suitable for moderately complex tasks. Its capabilities are always the same as the latest snapshot version. | Yes | |
Qwen-Long | qwen-long | The model in the Qwen series with the longest context window. It offers balanced capabilities at a low cost. It is suitable for tasks such as long-text analysis, information extraction, summarization, and classification. | No | |
Qwen-Long-Latest | qwen-long-latest | The model in the Qwen series with the longest context window. It offers balanced capabilities at a low cost. It is suitable for tasks such as long-text analysis, information extraction, summarization, and classification. Its capabilities are always the same as the latest snapshot version. | No | |
Qwen3-32b | qwen3-32b | An open-source Qwen model. It excels in inference, human preference alignment, agent capabilities, and multilingual abilities. It has 32 billion (32B) parameters. | Yes | |
Qwen3-235b-a22b | qwen3-235b-a22b | An open-source Qwen model. It excels in inference, human preference alignment, agent capabilities, and multilingual abilities. It has 235 billion (235B) parameters. | Yes | |
Qwen-Coder | qwen-coder-plus | The Qwen code model. | No | |
DeepSeek-R1 | deepseek-r1 | This high-performance version has powerful reasoning capabilities. It delivers strong performance on tasks such as mathematics, coding, and natural language inference. | Yes (deep thinking mode cannot be disabled) | |
DeepSeek-V3 | deepseek-v3 | A self-developed Mixture of Experts (MoE) model. It excels in long-text processing, coding, mathematics, encyclopedic knowledge, and Chinese language capabilities. | No | |
DeepSeek-V3.1 | deepseek-v3.1 | A 685B high-performance model released on August 20, 2025. It excels in long-text processing, coding, mathematics, encyclopedic knowledge, and Chinese language capabilities. | No | |
Kimi-K2 | Moonshot-Kimi-K2-Instruct | The Kimi series models are MoE language models from Moonshot AI. They show excellent performance in cutting-edge knowledge, inference, and coding tasks. | No | |
Qwen3-Next-80B-A3B(Thinking) | qwen3-next-80b-a3b-thinking | A new-generation open-source model based on Qwen3 that uses the thinking mode. Compared to the previous version (Qwen3-235B-A22B-Thinking-2507), it has improved instruction-following capabilities and provides more concise summaries. | Yes (thinking mode cannot be disabled) | |
Qwen3-Next-80B-A3B(Instruct) | qwen3-next-80b-a3b-instruct | A new-generation open-source model based on Qwen3 that uses the non-thinking mode. Compared to the previous version (Qwen3-235B-A22B-Instruct-2507), it has better Chinese text understanding, enhanced logical reasoning, and improved performance on text generation tasks. | No | |
Qwen3-235B-A22B(Thinking-2507) | qwen3-235b-a22b-thinking-2507 | A new-generation open-source model based on Qwen3 that uses the thinking mode. It is an upgraded version of qwen3-235b-a22b (thinking mode). | Yes (thinking mode cannot be disabled) | |
Qwen3-235B-A22B(Instruct-2507) | qwen3-235b-a22b-instruct-2507 | A new-generation open-source model based on Qwen3 that uses the non-thinking mode. It is an upgraded version of qwen3-235b-a22b (non-thinking mode). | No | |
DeepSeek | DeepSeek-thinking mode | deepseek-reasoner | This high-performance version has powerful reasoning capabilities. It delivers strong performance on tasks such as mathematics, coding, and natural language inference. | Yes (deep thinking mode cannot be disabled) |
DeepSeek-non-thinking mode | deepseek-chat | A self-developed MoE model. It excels in long-text processing, coding, mathematics, encyclopedic knowledge, and Chinese language capabilities. | No | |
Alibaba Cloud (AI Stack) | Qwen3-32B | Qwen3-32B | An open-source Qwen model. It excels in inference, human preference alignment, agent capabilities, and multilingual abilities. It has 32 billion (32B) parameters. | No |
Qwen3-235B-A22B-Instruct-2507 | Qwen3-235B-A22B-Instruct-2507 | A high-performance language model in the Qwen series designed for complex tasks. Released in July 2025, this model is an upgraded version of Qwen3-235B-A22B and supports the non-thinking mode. It excels in inference, general capabilities, and tool calling. It is suitable for scenarios that require high precision and complex logical processing. | No |