Models

Switch cloud or local models without changing the coding workflow.

List and select

silkcode models
silkcode --model auto ~/Projects/my-project
silkcode --model deepseek ~/Projects/my-project

--model auto prefers a detected local server, then a configured cloud provider.

Pull a local model

silkcode models pull qwen2.5-coder

Choose a model with tool-calling support because the agent reads, edits, tests, and manages repository state through tools.

Supported model families

Silk Code can work with DeepSeek, Qwen, Kimi, GLM, MiniMax, Cloudflare Workers AI, OpenRouter, Ollama, vLLM, LM Studio, and other OpenAI-compatible endpoints.