Supported LLMs

AI Assistant supports a variety of advanced cloud-based LLMs, as well as the option to use locally hosted models. This flexibility allows you to choose the most suitable model for your specific task. For example, you might want to use large models for complex codebase-related tasks, compact models for quick responses, or local models if you prefer to keep your data private.

Cloud-based models

The table below lists models available for selection in AI Assistant:

Model	Capabilities	Model context window
Claude 4.5 Sonnet		200k
Claude 4.5 Haiku		200k
Claude 4.1 Opus		200k
Claude 4 Sonnet		200k
Claude 3.7 Sonnet		200k
Claude 3.5 Sonnet		200k
Claude 3.5 Haiku		200k
Gemini 2.5 Pro		1M
Gemini 2.5 Flash		1M
Gemini 2.5 Flash-Lite		1M
Gemini 2.0 Flash		1M
GPT-5		400k^[1]
GPT-5 mini		400k^[1]
GPT-5 nano		400k^[1]
GPT-4.1		1M
GPT-4.1 mini		1M
GPT-4.1 nano		1M
GPT-4o		128k
o1		200k
o3		200k
o3-mini		200k
o4-mini		200k

Supported models history

The following table lists AI models that have been available in AI Assistant, along with the plugin versions in which they were introduced or removed.

Model	Added in version	Removed in version
Claude 4.5 Sonnet	2025.2.x	–
Claude 4.5 Haiku	2025.2.x	–
Claude 4.1 Opus	2025.2.x	–
Claude 4 Sonnet	251.26094.80.x	–
Claude 3.7 Sonnet	243.23654.270	–
Claude 3.5 Sonnet	243.23654.270	–
Claude 3.5 Haiku	243.23654.270	–
Gemini 2.5 Pro	251.23774.42.x	–
Gemini 2.5 Flash	251.23774.42.28.x	–
Gemini 2.5 Flash-Lite	251.26094.80.19	–
Gemini 2.0 Flash	243.23654.270	–
Gemini 1.5 Pro	2024.3	251.23774.42.28.x
Gemini 1.5 Flash	2024.3	251.23774.42.28.x
GPT-5	2025.2	–
GPT-5 mini	2025.2	–
GPT-5 nano	2025.2	–
GPT-4.1	251.23774.42.28.x	–
GPT-4.1 mini	251.23774.42.28.x	–
GPT-4.1 nano	251.23774.42.28.x	–
GPT-4o	2024.2	–
o4-mini	251.23774.42.28.x	–
o3	251.23774.42.28.x	–
o3-mini	243.23654.270	–
o1	243.23654.270	–
o1-mini	243.23654.270	251.23774.42.28.x

Local models

AI Assistant supports a selection of models available through Ollama, LM Studio, and other OpenAI-compatible endpoints like llama.cpp or LiteLLM. These models are optimized for local use, enabling powerful AI capabilities without the need for cloud access.

The default model context window for local models is set to 64 000 tokens. If needed, you can adjust this value in the settings.

For more information about setting up local models, refer to Use local models.

23 October 2025