The short answer
Choose Standard for routine syntax, short explanations, and narrow fixes. Choose Thinking when the answer must reconcile several constraints, trace a multi-step bug, or plan a careful refactor. Choose Max for the hardest architecture or debugging work where deeper reasoning is worth more time and credits. Standard is available on every plan, Thinking on Starter and above, and Max on Ultra. Modes are stable public controls over server-routed models; the underlying provider lineup can change without changing how you choose a mode.
Choose the lightest mode that fits the task
Start with the complexity of the reasoning, not the length of the prompt. A long file with one simple question can still fit Standard. A short question with several interacting constraints may justify Thinking. Max is most useful after you have provided strong evidence and a precise decision to make.
| Mode | Good fit | Tradeoff |
|---|---|---|
| Standard | Syntax, API reminders, small snippets, focused explanations, routine fixes | Fastest and lowest credit cost |
| Thinking | Multi-step debugging, competing constraints, careful refactors, design evaluation | More reasoning time and credits |
| Max | Hard architecture, ambiguous failures, high-impact decisions, the toughest investigations | Deepest reasoning, longest wait, highest credit cost |
Know which plans unlock each mode
The plugin shows the public modes and locks those above your current access level. Agent Mode is a separate plan capability: Pro can use Agent with Standard or Thinking, while Ultra can also choose Max. See the live Pricing page for current prices and monthly credit allotments instead of relying on copied numbers in an article.
| Plan | Available reasoning modes | Agent Mode |
|---|---|---|
| Free | Standard | Not included |
| Starter | Standard and Thinking | Not included |
| Pro | Standard and Thinking | Included |
| Ultra | Standard, Thinking, and Max | Included |
Match the same project question to different depth
@selection
Explain why this typed array assignment fails and show the smallest valid GDScript correction.@file res://scripts/inventory.gd
@scene res://ui/inventory.tscn --scripts
Trace why a moved item can appear in both slots. Reconcile the drag state, signal order, and UI refresh before proposing a fix.Evaluate whether this save-system boundary can survive scene reloads, versioned resources, and partial writes. Compare two designs against the attached constraints, identify migration risks, and recommend one with a verification plan.Keep in mind: A deeper mode cannot compensate for missing evidence. Attach the relevant error, script, scene, or reproduction before increasing reasoning depth.
Choose a mode, not a provider identifier
GDSense maps Standard, Thinking, and Max to a server-managed model lineup. The exact provider, deployment, or model identifier may change as routing, reliability, and product quality evolve. The public contract is the mode’s intended reasoning depth and credit cost, not a permanent raw model name.
This routing also means an old screenshot or response label should not be treated as a guarantee of the model used for a future request. Choose based on task difficulty and confirm the mode shown in the plugin before sending.
Control credit use without under-solving the task
Higher reasoning modes use more credits per request or completed Agent turn. Agent Mode does not add a separate multiplier; each completed turn uses the selected mode’s cost. The plugin shows the chosen mode, remaining balance, and completed Agent totals so you can decide whether another turn is worthwhile.
- Use Standard for the first focused pass when the task is routine.
- Improve the prompt and attachments before moving to a deeper mode.
- Use Thinking when several verified facts must be reconciled.
- Reserve Max for work where the additional depth materially changes the decision.
- Check the current balance and usage instead of assuming unused credits roll over.