Model choices
Which model to pick
Green Small is the default and the right choice for most tasks. Drop to Green Tiny for quick, simple questions; step up to Green Large when a task needs multi-step reasoning. Switch anytime from the model selector above the input area.
Which models you can use depends on your plan. Plan comparison and exact allocations are on the pricing page, and plans and limits explains how the token limits work.
The green models
There are three green chat models:
- Green Tiny: the same model as Green Small, with short, capped answers. Good for quick questions and simple tasks.
- Green Small: the everyday model and the default. Good for most work.
- Green Large: the strongest of the three. Good for detailed analysis and multi-step reasoning.
Green Tiny and Green Small run on certified renewable electricity at Infomaniak. Green Large does too, per GreenPT; its data center is not named.
Green Small and Green Large both offer a Reasoning switch behind the + button next to the input. On Green Small it turns step-by-step thinking on. Green Large always reasons, and the switch asks it to do more. Either way, answers are slower and use more energy.
Details on each model, including who makes it and where it runs, are on the about page.
The Claude models
Haiku and Sonnet, made by Anthropic, are available on the Advanced plan for work that needs stronger reasoning. Haiku is the faster of the two; Sonnet is the more capable.
Anthropic has not disclosed where Claude's electricity comes from. That is a compromise. Claude's energy is also estimated: Anthropic reports none. See measured vs estimated for how the estimates work.
What we leave out
No large frontier models, no image generation. That is on purpose. Smaller models handle most everyday tasks well. Whatever a model's capability, we exclude it if its maker scores below C, or gets any F, on the FLI AI Safety Index, or has dissolved its safety team.
If you need maximum capability regardless of cost, this may not be your tool.
The reasoning behind what we include is on the about page; the safety grades and the compromises we accept are on the transparency page.