Models and budgets

The budget is a length limit, not a quality setting: the same model writes equally well at every tier. The holds below are for a typical description. The builder counts your exact description and shows the real hold before you start, and a finished build is charged only for the tokens it used, often well below the hold.

Haiku 4.5

Fastest and cheapest

Best for: Small plugins. A handful of commands, simple storage, quick retries while you iterate on the description.

Trade-off: It writes straight out without reasoning first, so intricate logic and long command lists are where it slips.

Budget	Room for one pass	Hold for a typical build
Standard	32,768 tokens	$0.51
Large	48,384 tokens	$0.75
Max	64,000 tokens	$0.98

Sonnet 5.5

The balanced defaultReasons before writing

Best for: Most plugins. Several commands, a dashboard page, real permission checks.

Trade-off: Reasons before writing, so a build takes longer than Haiku and those reasoning tokens are billed.

Budget	Room for one pass	Hold for a typical build
Standard	32,768 tokens	$1.03
Large	80,384 tokens	$2.45
Max	128,000 tokens	$3.88

Opus 5.5

For complex pluginsReasons before writing

Best for: Many interacting commands, non-obvious rules, background jobs, anything you expect to get subtly wrong.

Trade-off: Costs more per token than Sonnet for work Sonnet often handles.

Budget	Room for one pass	Hold for a typical build
Standard	32,768 tokens	$2.05
Large	80,384 tokens	$4.91
Max	128,000 tokens	$7.76

Fable 5.1

The most capable modelReasons before writing

Best for: The hardest builds. Deep reasoning over a long, detailed spec that the other rungs keep getting wrong.

Trade-off: The most expensive rung by a wide margin, and the slowest to finish. Reach for it when Opus has already failed you.

Budget	Room for one pass	Hold for a typical build
Standard	32,768 tokens	$5.13
Large	80,384 tokens	$12.27
Max	128,000 tokens	$19.41

YourBot docs Reference tables are generated from the code that is running. Ask in Discord Suggest a correction