Skip to Content
GuideChoosing a Model

Choosing a Model

Agent8 supports multiple AI models. You can select the model from the dropdown in the prompt bar before running, or change it at any time during a session.

Available Models

ModelDescription
Agent8 AutoAutomatically selects the best model for each task. Recommended for most users.
Claude Sonnet 4.6Anthropic’s fast, capable model — strong at following instructions and structured code generation
Claude Opus 4.6Anthropic’s most powerful model — best for complex, multi-file games requiring deep reasoning. Note: Credit consumption can be very high.
GPT-5.4OpenAI’s latest model — strong overall performance
Gemini 3.1 ProGoogle’s high-performance model — good for creative and visual tasks
Gemini 3 ProGoogle’s capable mid-tier model
Gemini 3 FlashGoogle’s fast model — lower cost per generation
Grok 4xAI’s model — strong at reasoning and code

New models are added to Agent8 as they become available. Check the model dropdown for the latest options.

Which Model Should I Use?

  • Agent8 Auto — If you’re unsure, start here. It picks the right model automatically.
  • Faster / cheaper — Gemini 3 Flash for quick iterations
  • Best quality — Claude Opus 4.6 or GPT-5.4 for complex games
  • Specific preference — Choose the model you’re most familiar with

Model Cost

Each AI response consumes credits. The cost varies by model and the length of the conversation. You can track usage in Logs and check your balance in Credits.

Last updated on