Models
It runs on all of them, and picks
DuoHand is tied to no single model company. It keeps a shelf and chooses per step: the sharpest model to argue about a direction, a fast one to answer a roll call, the right one to render.
Thinking
Language models on the shelf
Anthropic
Claude Sonnet 5, Opus 5 and Fable 5.1.
OpenAI
GPT-5.6 Sol and GPT-6 Astra.
Google
Gemini 3.8 Flash.
DeepSeek
DeepSeek V4-Pro and V4.1-Flash.
Z.ai
GLM-5.3 and GLM-5.3 Flash.
Whatever ships next
A new model joins the shelf without you migrating anything.
Pictures
Image models on the shelf
OpenAI
GPT Image 2.5, in two cuts.
Google
Gemini 3 Pro Image.
Qwen and Black Forest
Qwen-Image Max and FLUX.2 [pro].
Questions
About the shelf
Do I have to choose?
No. The default picks per step; you can pin one when you want a particular voice.
Does a sharper model cost more?
Work costs what it costs, and every answer carries its price before you spend again.
What if one is retired?
The shelf is checked every week. A retired model stops being offered rather than failing mid-job.
Try the same question twice
On two models, and compare.