← All models

Muse Wonder

Language Preview Full comparison ↗

meta/muse-wonder-1.0 · by Meta · decoder-only-transformer

Pricing — 1 offering(s)

Meta

About this offering: Meta's most capable model, for complex software engineering, long-running agents and computer use. In preview on the Meta Model API (pay-as-you-go, Standard mode only), with a 1,048,576-token context window and no long-context premium. Muse Wonder Full, the higher-capability variant, is only available through a Muse Code subscription and has no per-token price.

Input tokens

  • $4.00 / 1M tokens (input) Current 2026-10-13 → present

Output tokens

  • $20.00 / 1M tokens (output) Current 2026-10-13 → present

    Reasoning tokens are billed as output.

Cached input tokens

  • $0.20 / 1M tokens (cached input) Current 2026-10-13 → present

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

See how Muse Wonder fits into a cost-aware routing setup

See how →

Capability profile

How Muse Wonder rates across core capability dimensions, with the task-level evidence behind each rating.

reasoning Strong

Reasoning effort from minimal to max; Meta reports its benchmark results at max effort.

coding Strong

Meta's model page: "Solve complex engineering tasks in large repositories, making and verifying changes across multiple files in parallel." Vendor-reported; no independent coding benchmark recorded yet.

tool use Strong

Meta's model page: plans long tasks, delegates to sub-agents, calls tools in parallel and recovers from errors; native computer use for browsers, desktop and mobile apps.

instruction following Strong
context window Strong

1,048,576-token context window, with no long-context price premium.

multilingual Strong

Meta's model card: trained on multilingual data with substantial representation across worldwide languages.

cost efficiency Moderate

$4/$20 per 1M tokens, with cached input at $0.20; the same as Claude Opus 5.5 and well below GPT-6 Astra's $10/$50.

Ratings are estimated — limited independent data is available for this model.

Operator guidance

Route here for long-horizon agentic and coding work where Meta's most capable pay-as-you-go model is wanted. It's priced like Claude Opus 5.5, so compare on task results rather than price. For high-volume agentic work, Muse Spark 1.3 costs about a third as much for input and a fifth for output. The model is in preview.

Use cases

Limitations

Citations