The model picker in the prompt composer chooses which AI builds for you. Three models are available to everyone:

​
Muranyi Fast — cheap, quick edits

The iteration specialist: small fixes, tweaks, and focused changes on an existing game, noticeably faster and at a fraction of the credits of a full Muranyi 3.5 build.
  • Rate: ~110 credits per million AI tokens (blended) — see the exact rate table. Its output rate is 2.5× cheaper than Muranyi 3.5, but most of the saving comes from doing less work per prompt: a focused edit rather than a whole build.
  • Typical edit: 35–65 credits.
  • Edit-only: Muranyi Fast can’t start new games — it appears in the picker once your game exists.
  • Use it for: “make the coins spin”, “raise the jump height”, small bug fixes — any change where you know exactly what you want.

​
Muranyi 3.5 — the balanced default

A dependable all-rounder for building complete games: complex ideas, larger worlds, full game logic, at a balanced cost. This is the right choice for many prompts.
  • Rate: ~95 credits per million AI tokens (blended) — the standard rate.
  • Typical prompt: 45–550 credits.
  • Use it for: everyday building, iteration, bug fixes, and simpler new games.

​
Atomos — the most capable

Tesana’s smartest, most creative model — best-in-class for ambitious games and complex logic. It often nails in one prompt what can take 10+ prompts on other models.
  • Rate: 1.67× Muranyi 3.5 per token (~132 credits per million AI tokens blended). It costs more per prompt than that ratio suggests, because it also does far more work per prompt.
  • Typical prompt: 275–3,600 credits.
  • Requires Max or Ultra. On Pro and below the picker shows it locked — a Pro month of credits does not go far enough on Atomos to be worth selling.
  • Use it for: ambitious new games, intricate systems (simulation, AI opponents, procedural generation), prompts where Muranyi has plateaued.
Exact per-token-type rates for all three models: Pricing.

​
Picking and switching

  • The picker lives in the prompt composer on the Dashboard and in the Studio. Your choice persists per surface (and per game in the Studio).
  • You can switch any time, per prompt. A common pattern: start the game on Atomos to get a strong foundation, then iterate on Muranyi 3.5 to stretch your credits — or build on Muranyi and bring in Atomos for the prompts that need extra firepower.
  • The model badge on each ledger row in your Usage page shows which model ran each prompt.
A higher rate does not mean a higher cost for the same outcome. On hard prompts Atomos is often cheaper overall, because it finishes in one prompt what a weaker model needs 3+ prompts for. Count the whole exchange, not the single prompt.

​
What your plan changes, and what it doesn’t

Rates never change. A prompt costs the same on a free account as on Ultra — the per-token rates are one published table for everyone, with no volume curve and no plan discount. Two things are plan-gated, and the picker shows them locked rather than failing later: Everything else — Muranyi Fast, Muranyi 3.5, switching per prompt — is available on every plan, including the free daily allowance.