Atom Config
The Atom Config is a quick-access menu that lets you switch your model and thinking mode without opening Settings.
Click the Atom orb in the bottom-left of the chat input to open it.
What’s in the menu
Provider & Model
The top of the Atom Config shows two dropdowns:
- Provider — Use the built-in OpenAI experience, or choose Anthropic if you set up Claude Code.
- Model — Choose the model for the selected provider.
Changes take effect on your next message. If Atom detects a compatibility issue with your selected provider, a notice appears below the dropdowns.
Thinking Modes
The bottom row controls how much reasoning Atom applies to your requests. Your choice persists across messages until you change it.
Quick
Minimal reasoning, optimized for speed and cost. Best for straightforward, repetitive tasks.
When it’s useful:
- Simple batch operations (rename layers, reset properties)
- Quick property adjustments
- Tasks you’ve done before in the same session
Tradeoff: Faster and cheaper, but may miss nuance on complex requests.
Balanced (Default)
Solid reasoning for moderately complex tasks.
When it’s useful:
- Multi-step workflows with dependencies
- Rigging and control setups
- Expression creation with specific logic
Tradeoff: Slightly slower than Quick, but handles complexity better.
Deep
Extended thinking for challenging or ambiguous requests.
When it’s useful:
- Complex animation sequences with timing constraints
- Intricate expression logic across multiple layers
- Tasks requiring careful analysis of layer relationships
Tradeoff: Slower and more expensive, but solves harder problems more reliably.
Cost and Speed
Higher thinking modes use more compute, which means:
- Longer response times — Deeper reasoning takes more time
- Higher token usage — More reasoning consumes more tokens from your AI provider
For most work, Balanced gives you the best mix. Switch to Quick for simple repetitive tasks or Deep when you need maximum reasoning power.