This version is still in development and is not considered stable yet. For the latest stable version, please use Spring AI 2.1.0!

Chat Model Comparison

The table compares the chat model integrations of Spring AI:

  • Input: what the model accepts besides text, see Multimodality.

  • Tool calling: whether the model can call tools.

  • Streaming: whether the model can stream its answer.

  • Retry: whether failed calls are retried, see Retry.

  • Observability: whether the model reports metrics and traces, see Observability.

  • Native structured output: whether the chat options support provider-native structured output.

  • Local: whether the model can run on your own machine.

  • OpenAI-compatible: whether the service also offers an OpenAI-compatible API.

Provider Input Tool calling Streaming Retry Observability Native structured output Local OpenAI-compatible

Anthropic

text, PDF, image

Yes

Yes

Yes

Yes

Yes

No

No

Amazon Bedrock Converse

text, image, video, documents

Yes

Yes

Yes

Yes

Yes

No

No

DeepSeek

text

Yes

Yes

Yes

Yes

No

No

Yes

Google GenAI

text, PDF, image, audio, video

Yes

Yes

Yes

Yes

Yes

No

No

Mistral AI

text, image

Yes

Yes

Yes

Yes

Yes

No

Yes

Ollama

text, image

Yes

Yes

Yes

Yes

Yes

Yes

Yes

OpenAI Chat Completions

text, image, audio; answers with text or audio

Yes

Yes

Yes

Yes

Yes

No

Yes

OpenAI Responses API

text, image, PDF; answers with text or images

Yes

Yes

Yes

Yes

Yes

No

-

MiniMax, through the Anthropic module

text, image

Yes

Yes

Yes

Yes

Depends on the service

No

Yes

Groq, through the OpenAI module

text, image

Yes

Yes

Yes

Yes

Depends on the service

No

Yes

NVIDIA, through the OpenAI module

text, image

Yes

Yes

Yes

Yes

Depends on the service

No

Yes

Perplexity, through the OpenAI module

text

No

Yes

Yes

Yes

Depends on the service

No

Yes

Docker Model Runner, through the OpenAI module

depends on the model

Depends on the model

Yes

Yes

Yes

Depends on the model

Yes

Yes

Qxotic jinfer (Community)

text, image, audio, video, depending on the model

Yes

Yes

No

Yes

Yes

Yes

No

Services that are used through another module, such as the OpenAI module, use the chat options of that module. Whether they honor native structured output, or an option of the original provider, depends on the service.