Skip to Content
🇬🇧Models

Alex Models and Limits

Alex supports multiple AI models for different kinds of work. Some are optimized for quick answers, others for coding, deep analysis, or complex decisions. The selected model affects response quality, speed, and credit consumption.

Exact model availability depends on your tariff and the current service configuration. In chat, you only see models available to your account.


How Models Affect Consumption

Consumption is mainly affected by:

FactorImpact
Model complexityStronger models consume more credits but handle complex tasks better
Conversation lengthA long thread may require more context
Amount of loaded dataLogs, files, command output, or web content increase the amount of work
Tools usedDiagnostics, sub-agents, and multi-step actions can add consumption
CacheRepeated context can be processed more efficiently on supported models

Alex calculates consumption automatically. The panel shows your limit state and consumption history without requiring you to manage technical details.


Model Groups

Standard model

The standard model is suitable for everyday use:

  • quick questions,
  • server status checks,
  • simple configuration,
  • error explanations,
  • regular file edits,
  • continuing from an existing plan.

For most tasks, this is the best starting point because it balances quality, speed, and consumption.

Coding and analytical models

These models are useful when Alex works with code, logs, or multiple files:

  • application debugging,
  • configuration review,
  • refactoring,
  • root cause analysis,
  • comparing multiple solution options.

Consumption can be higher especially when the task requires a larger amount of context.

Premium models

Premium models are best for complex or higher-risk situations:

  • production incidents,
  • architecture decisions,
  • larger migrations,
  • deep debugging,
  • security review,
  • problems where the standard model repeatedly does not provide a good enough result.

Premium models usually consume more credits. Use them where higher quality is worth the additional consumption.


Cache on Models

Some models support prompt cache. If Alex repeatedly works with the same or similar context, part of that context may be processed more efficiently.

Cache helps most when you continue:

  • in the same conversation,
  • on the same server or project,
  • after a previous log analysis,
  • from an existing plan,
  • through iterations on the same files.

For best results, continue in the same thread and avoid resending content Alex has already read.


When To Use Which Model

SituationRecommendation
Quick question or explanationStandard model
Simple server managementStandard model
Installation or configuration with clear instructionsStandard or coding model
Debugging an error from logsCoding or analytical model
Unclear incident with unknown causePlanning Mode + suitable stronger model
Architecture, migration, securityPremium model when available

Tip: Start with the standard model. If Alex reaches a complex part of the task, switch to a stronger model only for that part.


How To Save Credits When Choosing Models

  1. Do not use premium models for routine questions. Stronger models are most useful for complex work.
  2. Use Planning Mode before large changes. Get the approach first, then execute actions.
  3. Keep one thread per problem. This helps context and cache.
  4. State the goal clearly. Fewer corrective follow-ups means less unnecessary consumption.
  5. After a long analysis, name the next step. For example: “Continue from the plan and execute only step 1.”

Limit Indicator in Chat

The chat shows your limit state and warnings when you are getting close to exhaustion. Colors provide a quick orientation:

ColorState
GreenEnough room for more work
YellowConsumption is growing; consider task scope
OrangeYou are getting close to the limit
RedLimit is exhausted or very close

When a limit is exhausted, Alex shows the next available options.


Switching Models

  1. In chat, click the settings icon in the header.
  2. Select the desired model.
  3. The change applies to future messages.

If you are unsure, keep the default model. Alex is designed to work well for regular tasks without manual tuning.


Next Steps


Need help choosing a model? Open a support ticket  or ask Alex in your panel.

Last updated on