Help Center/Code & engines
Choosing a model in Prompt to Browser Game
Which Anthropic model to pick for browser-game generation, what each costs, and when to use each.
Prompt to Browser Game supports five Anthropic models, each with its own credit cost and strengths. The picker sits at the top of the tool page alongside the streaming and Advanced toggles.
Claude Haiku 4.5 (5 credits quick / 15 full, all plans): the cheapest and fastest option. Perfect for game jams, prototyping mechanics, or exploring an idea before committing.
Claude Sonnet 5 (20 / 60, all plans, default): the right balance of quality and speed for most games. If you are unsure, start here.
Claude Sonnet 4.6 (15 / 45, all plans): the previous default, retained for backwards compatibility. Slightly cheaper than Sonnet 5 but produces less polished output on complex briefs.
Claude Fable 5 (25 / 75, Indie plan or higher): tuned for creative writing. Reach for it when the game is narrative-heavy — dialogue-driven, story-forward, unusual mechanics that need imaginative framing.
Claude Opus 5.5 (100 / 300, Indie plan or higher): frontier reasoning. The strongest model for ambitious multi-mechanic games with intricate rule sets, but 5-10x the cost of Sonnet.
"Quick" and "Full" refer to the pipeline, not the model. Quick is one Claude call; Full is three passes (design, skeleton, implementation) with the model doing dedicated work at each stage.
Tip
If you have your own Anthropic API key, add it at /dashboard/api-keys and tick "Use my Anthropic key" — every generation then costs just 1 GameFoundry credit regardless of the model, and Anthropic bills your account directly.