Choose a model for the work
A practical way to compare models, understand our named configurations and keep the choice tied to your actual task.

On this page
In brief
Choose for the task you can describe, then check the available model and credit estimate before sending.
Weaver, Swift and Owl are TwelveBirds configurations of external provider models, not TwelveBirds-trained foundation models.
Compare models with the same brief and sources. Another response uses credits, so decide what the comparison should teach you.
Start with the task, then the model
Choosing a model becomes easier when you can describe the work. A short rewrite needs a different review from a comparison of several options. An image prompt needs something different again. Before opening a model picker, decide what a useful answer would contain and how much detail you actually need.
In Twelva, the model picker lets you browse the options available to your workspace. Read the descriptions and review the credit estimate for your request. Availability and the catalog can change, so the current picker is a better reference than a model name remembered from an older article.
What our named configurations mean
Weaver, Swift and Owl are TwelveBirds configurations built on models from external providers. They are not foundation models trained by TwelveBirds. The configurations combine provider models with different instructions and settings, and the picker identifies the underlying model.
Their intended directions are straightforward: Weaver focuses on creative work and websites, Swift favors brief, direct answers, and Owl is oriented toward questions that need more consideration of trade-offs. These descriptions are starting points, not promises that one option will always outperform another. You can also select an available provider model directly.
Compare something you actually need
Use a small, representative task and keep the instructions and source material the same. For a product description, compare factual accuracy, whether it follows your requested voice, and how much editing remains. For a decision, look for explicit assumptions and a clear account of the trade-offs.
Twelva offers Change model below a completed response so you can request another answer with a different model. That is a new response and uses credits. Decide in advance what the comparison should teach you; repeated generations without a review standard can create more options without helping you choose.
| Your task | Selection criterion | What to check |
|---|---|---|
| A short rewrite | A brief, direct response that follows your requested voice. | Are the original facts intact? Does the result fit the requested length? |
| A creative draft | A usable direction that follows your brief and brand choices. | Are the claims supported? What needs editing before you can use it? |
| A decision with trade-offs | Explicit assumptions and a clear comparison of the options. | Are the assumptions valid? Do the sources support the recommendation? |
Keep the choice visible
Save useful models as favorites, and consider a project default when the same kind of work comes back regularly. Selecting a project can change the current chat model, so check the picker before sending when the choice matters. Choosing a model for a future message does not rewrite the answers already in the conversation.
Image and video generation have their own tool and model choices in the creation cards within Chat. A chat model that accepts an image is not the same thing as an image-generation tool. Start with the output you need, use the controls available for that task, and judge the result against your own brief.



