Choosing a model
Buildprint lets you use Buildprint-managed models at cost, or connect your ChatGPT / Claude subscription to use their limits.
Recommended setup
It is strongly recommended to connect either a ChatGPT or Claude subscription to Buildprint. This will allow you to use Buildprint at the lowest cost, as it will use your existing subscription. Note that each team member will be required to connect their own ChatGPT/Claude subscription - it is not shared among the workspace.
Understanding your app
For Read only interactions, any model can be used. Minimax M2.7 is the cheapest Buildprint managed model.
Opus with Claude will consume your rate limits faster, so if you're not on the Claude Max plan, it's recommended to use Sonnet for understanding the app.
For ChatGPT users, it is recommended to use GPT 5.3 Codex or GPT 5.4 for all tasks.
Editing your app
For editing your app, you must use GPT 5.4, GPT 5.3 Codex, or Opus 4.6. These are the models which demonstrate the best ability to use Buildprint effectively.
For front-end work, Opus 4.6 is marginally better than the GPT family. For complex logic, the GPT family of models is more effective. However, there is not much in it.
Buildprint-managed models
Buildprint managed models allow you to use other AI models at cost. You can top up credits in the workspace settings. Currently, the cheapest model is Minimax M2.7 and this is recommended for understanding your app. The only model available for editing is GPT 5.3 Codex.
As model's intelligence per cost increases, Buildprint will become cheaper to use on a pay-as-you-go basis, as well as faster.
Cost management
Buildprint (and AI in general) is not cheap. If cost is a concern for you, then it's important to manage your conversations and context appropriately (especially when using pay-as-you-go Buildprint credits). Keep conversations short, and provide specific guidance in your prompt. The more specific you are, the less searching Buildprint has to do. You can use the cheapest model for your use case (e.g understanding an app requires a less powerful model than editing an app). You can stop your conversation at any time if you think the AI is getting distracted, and point it in the right direction.