What are the supported models in Google ADK? Explain all model providers and how to use them.
#google-adk#models#gemini#openai#anthropic#ollama#litellm
Answer
All Supported Models in Google ADK
Google ADK supports 100+ LLMs across 8 model providers, from Google Gemini to local Ollama models.
Model Provider Architecture
All Model Providers
| # | Provider | Models | Connection | Docs |
|---|---|---|---|---|
| 1 | Google Gemini | gemini-2.5-flash, gemini-2.5-pro | Direct | Link |
| 2 | Anthropic | Claude Sonnet, Haiku, Opus | Vertex AI + Registry | Link |
| 3 | OpenAI | GPT-4o, GPT-4-Turbo | LiteLLM | Link |
| 4 | Vertex AI | Llama 3, fine-tuned models | Vertex endpoint | Link |
| 5 | Ollama | Gemma, Llama, Mistral (local) | LiteLLM + Ollama | Link |
| 6 | vLLM | Any self-hosted model | LiteLLM + vLLM | Link |
| 7 | LiteRT-LM | On-device edge models | Local server | Link |
| 8 | Apigee | Enterprise-governed models | AI Gateway | Link |
Code Examples
Gemini (Direct)
pythonagent = Agent(name="a", model="gemini-2.5-flash", instruction="...")
OpenAI (LiteLLM)
pythonfrom google.adk.models.lite_llm import LiteLlm agent = Agent(name="a", model=LiteLlm(model="openai/gpt-4o"), instruction="...")
Anthropic (Registry)
pythonfrom google.adk.models.anthropic_llm import Claude from google.adk.models.registry import LLMRegistry LLMRegistry.register(Claude) agent = Agent(name="a", model="claude-sonnet-4-6", instruction="...")
Ollama (Local)
pythonimport os os.environ["OLLAMA_API_BASE"] = "http://localhost:11434" from google.adk.models.lite_llm import LiteLlm agent = Agent(name="a", model=LiteLlm(model="ollama_chat/gemma3:latest"), instruction="...")
vLLM (Self-hosted)
pythonfrom google.adk.models.lite_llm import LiteLlm agent = Agent(name="a", model=LiteLlm(model="hosted_vllm/google/gemma-3-4b-it"), instruction="...")
Model Feature Support
| Feature | Gemini | OpenAI | Claude | Ollama |
|---|---|---|---|---|
| Tool calling | Full | Full | Full | Partial |
| Streaming | Full + Live | SSE | SSE | SSE |
| Code execution | Built-in | No | No | No |
| Grounding (Search) | Yes | No | No | No |
| Vision (images) | Yes | Yes | Yes | Model-dependent |
| Audio I/O | Yes (Live) | Via Whisper | No | No |
Pro tip: Start with Gemini for full feature access, then expand to other models as needed.
Learn more at Models Overview.