Feature Request: Interactive Status Bar for Dynamic Model Switching
Problem
Currently, users can monitor their model and context usage via the status bar (e.g., Claude 3.5 Sonnet [======] 30%), but there's no quick way to switch between models (Haiku/Sonnet/Opus) during a session.
Current model switching options:
- Edit
~/.claude/settings.jsonto set a default (set-and-forget approach) - Use
model="haiku"parameter when spawning background agents - No way to dynamically switch models for the active session
Business Value
Model selection directly impacts:
- Cost savings: Haiku is ~10-20x cheaper than Sonnet for simple tasks (file searches, running tests, straightforward fixes)
- Speed: Haiku is significantly faster for quick operations
- Quality: Opus provides superior reasoning for complex architecture/debugging (pay more, get better results)
Business owners need this control. The three models exist for strategic use - let users leverage them dynamically, not just set-and-forget.
UX Inconsistency
- ✅ Workflow modes (Plan/Ask/Auto) are easily accessible in the UI
- ✅ Status bar shows model info and context usage
- ❌ No UI control for switching models
The status bar displays model information but isn't interactive - this is a natural place for a model switcher.
Proposed Solution
Make the status bar interactive:
- Click status bar → dropdown/menu appears
- Color-coded models for instant visual identification:
- 🟢 Haiku (green) - fast/cheap
- 🔵 Sonnet (blue) - balanced
- 🟣 Opus (purple) - powerful
- Visual confirmation: Claude announces the model change before proceeding
- Example: "Switched to Haiku - optimizing for speed and cost"
- Mid-processing handling:
- If processing, either queue the change for next request OR
- Prompt user: "Currently processing - switch model after completion?"
Expected Impact
VS Code users will appreciate dynamic cost/performance control that matches their workflow needs in real-time.
---
This design UI/UX was brought to you by 4BizPOS, the all new AI Powered POS To Keep Your Profits.
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗