Dify supports the below model providers out-of-box:

ProviderLLMText EmbeddingRerankSpeech to textTTS
OpenAI✔️(🛠️)(👓)✔️✔️✔️
Anthropic✔️(🛠️)
Azure OpenAI✔️(🛠️)(👓)✔️✔️✔️
Gemini✔️
Google Cloud✔️(👓)✔️
Nvidia API Catalog✔️✔️✔️
Nvidia NIM✔️
Nvidia Triton Inference Server✔️
AWS Bedrock✔️✔️
OpenRouter✔️
Cohere✔️✔️✔️
together.ai✔️
Ollama✔️✔️
Mistral AI✔️
groqcloud✔️
Replicate✔️✔️
Hugging Face✔️✔️
Xorbits inference✔️✔️✔️✔️✔️
Zhipu AI✔️(🛠️)(👓)✔️
Baichuan✔️✔️
Spark✔️
Minimax✔️(🛠️)✔️
Tongyi✔️✔️✔️
Wenxin✔️✔️
Moonshot AI✔️(🛠️)
Tencent Cloud✔️
Stepfun✔️(🛠️)(👓)
VolcanoEngine✔️✔️
01.AI✔️
360 Zhinao✔️
Azure AI Studio✔️✔️
deepseek✔️(🛠️)
Tencent Hunyuan✔️
SILICONFLOW✔️✔️
Jina AI✔️✔️
ChatGLM✔️
Xinference✔️(🛠️)(👓)✔️✔️
OpenLLM✔️✔️
LocalAI✔️✔️✔️✔️
OpenAI API-Compatible✔️✔️✔️
PerfXCloud✔️✔️
Lepton AI✔️
novita.ai✔️
Amazon Sagemaker✔️✔️✔️
Text Embedding Inference✔️✔️
GPUStack✔️(🛠️)(👓)✔️✔️

where (🛠️) ︎ denotes “function calling” and (👓) denotes “support for vision”.


This table is continuously updated. We also keep track of model providers requested by community members here. If you’d like to see a model provider not listed above, please consider contributing by making a PR. To learn more, check out our contribution.md Guide.