Model downloads

A recipe is one model at one precision. snowllm pull fetches it:

snowllm recipes                      # what there is
snowllm pull qwen3.6-35b-a3b-fp8     # or just `snowllm pull`, and pick from the list

• Fetches over several connections at once, and resumes where it stopped
• Verifies SHA-256 for files whose repository or catalogue provides a digest
• A recipe that wants a newer snowllm downloads anyway and says which version it wants

Checkpoints live in ~/models by default. Set models_dir in the JSON configuration or override it with SNOWLLM_MODELS. snowllm <recipe> looks there; snowllm pull --dir PATH downloads directly into that checkpoint directory; it does not change the configured model root. Serve that path directly if it is outside the root.

See Configuration for catalogue URLs, mirrors and download credentials.

Or fetch it yourself. Use a local checkpoint directory with a supported architecture and weight format:

hf download Qwen/Qwen3.6-35B-A3B-FP8 --local-dir ~/models/Qwen3.6-35B-A3B-FP8

A GGUF works too — snowllm pull qwen3.6-35b-a3b-q4-k-xl-dflash gets the UD-Q4_K_XL quant with its DFlash drafter beside it: a smaller download, and llama.cpp serves the same file.

Choosing a model

See supported models and resident weight sizes. Leave room for workspaces and caches beyond the listed weight memory.

For custom fine-tuned checkpoints, see the FAQ.