Models

Akio uses a curated whitelist of models to ensure every supported model works reliably with all features — including tools, MCP, orchestration and more. Rather than allowing arbitrary models, whitelisting lets us guarantee compatibility and a consistent experience across the board.

All models are hosted on Hugging Face and served in GGUF format.

Model	Repo	Parameters	Tags
Qwen3	Fastiraz/Qwen3	0.6B, 1.7B, 4B, 8B, 14B, 32B	llm
Qwen3.5	Fastiraz/Qwen3.5	9B	llm
Qwen3.6	Fastiraz/Qwen3.6	27B, 35B	llm, mtp
Qwen3.8	Fastiraz/Qwen3.8	27B	llm
Ornith	Fastiraz/Ornith-10	9B, 35B	llm
Ornith-1.5	Fastiraz/Ornith-15	9B, 35B, 397B	llm
Qwythos	Fastiraz/Qwythos	9B, 27B	llm
DeepSeek-R1	Fastiraz/DeepSeek-R1	1.5B, 7B, 14B, 32B	llm
Qwen3-Embedding	Fastiraz/Qwen3-Embedding	0.6B, 4B, 8B	embedding
Z-Image-Turbo	Fastiraz/Z-Image-Turbo	6B	image

Start with a smaller model like ggml-org/Qwen3-0.6B-GGUF for faster testing on CPU. You can also start with the Z-Image-Turbo model for the fastest image generation testing.