Different models trade off speed, cost, reasoning, visual understanding, and coding depth. Squid shows the selected model and expected credit range before generation so the cost decision is visible before work starts.
Choose for the task
- Use a starter model for focused apps, contained edits, and early product exploration.
- Use a multimodal-capable model when screenshot interpretation is central to the task.
- Use a stronger coding or reasoning model for complex state, dense interfaces, or difficult repairs.
- Keep the prompt and acceptance criteria constant when comparing two models.
Available models and their prices can change as providers evolve. Treat the model selector and its current estimate as the source of truth at the moment you generate.
Credit visibility
Squid’s generation flow separates the estimate shown before work from the usage recorded after a successful result is saved. The usage ledger helps you review project activity and credit changes over time.
If a request fails or a charge looks unexpected, preserve the project, model, time, and visible error details before contacting support.
Reduce avoidable retries
Credits are best spent on clear, evaluable work:
- Start with one primary product job.
- Include important behavior and states in the first brief.
- Repair correctness before requesting visual polish.
- Make focused follow-ups instead of repeatedly regenerating the full app.
- Preserve working areas explicitly during edits.