Browse by feature
Filter open-source AI models by capability.
Tool calling
10 models
Native structured function-call output. The foundation for reliable agent workflows.
JSON mode
10 models
Guaranteed valid-JSON output — either natively trained or via provider grammar-constraint layer.
Vision
5 models
Multimodal models with a vision encoder that can read images alongside text.
Long context
30 models
Context window of 100K tokens or more. For book-length inputs and monorepo-scale code contexts.
Reasoning
7 models
Trained to generate long internal chains-of-thought before answering. Higher latency, higher accuracy on hard problems.