Build quality levels
#| Level | What the app says | Default model | Notes |
|---|---|---|---|
| Economy | "Good for simple edits, fixes, and basic pages." | Gemini 3 Flash | Marked Default: new chats start here until you set another default. The only level on the Free plan |
| Standard | "Beautiful, detailed design. Cleanest results, fewest errors." | Claude Opus 5.5 | Pro and Max |
| Ultra | "Our most capable model. For the hardest, highest-stakes builds." | Claude Fable 5 | Marked Priciest. The app warns that it "may use credits several times faster than Economy or Standard". Pro and Max |
Switching to Economy or Ultra asks you to confirm first. The Economy warning says: "On complex tasks, it can sometimes repeat or get stuck in a loop. If that happens, press Stop to avoid unnecessary credit use."
Models you can choose
#Each level has a short list of models. The default model is listed first.
| Level | Models in the picker |
|---|---|
| Economy | Gemini 3 Flash, GLM 5.3 Flash, DeepSeek V4.1 Flash, GPT-6 Luna |
| Standard | Claude Opus 5.5, GPT-6.1 Sol, Claude Sonnet 5.5, Grok 4.7, Gemini 3.8 Flash, Claude Opus 4.6, Kimi K3 |
| Ultra | Claude Fable 5, Claude Fable 5.1, GPT-6 Astra |
Models join and leave these lists often. The picker always shows the current list, with each model's cost as one to four dots and a word: Lowest cost, Mid cost, High cost or Highest cost. The picker shows no price per message.
Every model in these lists can look at pictures you attach. Some cannot open PDF files: with those, the agent asks you for the text or a screenshot of the pages.
Effort and Context
#| Dial | Levels | Default | What it changes |
|---|---|---|---|
| Effort | Low, Medium, High, Extra High, Max | Extra High on Standard and Ultra. On Economy, each model keeps its own setting until you choose a level | How hard the model thinks on each step. The app says: "higher is better on complex builds but slower." Thinking is billed like the rest of the model's output, so a lower level spends fewer credits on thinking |
| Context | Low, Medium, High, Extra High, Max | High | How much of the conversation the agent keeps before it summarizes older messages. The app says: "Larger context can cost more, and on some models context above ~200K tokens costs roughly 2x." Choosing Extra High or Max asks you to confirm |
The Free plan shows Context but not Effort.
Where to change them
#| What | Where |
|---|---|
| This chat's level, model, Effort and Context | Choose the level button in the message box's toolbar (it shows the current level, like Economy). Choose build quality opens: choose a card for the level, or the model name on a card to pick another model. Effort and Context are below the cards |
| The starting point for new chats | Set as my default at the bottom of Choose build quality. The app says: "New chats start from your default; this chat stays as set." |
| All defaults in one place | Settings, AI defaults (under You): Default quality, Default effort and Default context |
| What this chat has spent so far | Open the project menu (☰ and the project name): the bottom shows Spent in this chat |
| What the workspace spent | Settings, Usage |
Background tasks
#Background tasks never run above the chat's level. Tasks that only read the project or research the web run on Economy; tasks that write or review code run at the chat's level. See How building with the agent works.