Independent guide · Checked 16 October 2026

Sources & calculation method

Every price needs a source. Every estimate needs assumptions. Here is what we verified, what we calculated and what remains unverified.

Primary sources

Last editorial check: 16 October 2026. Follow the current provider page and your regional checkout before spending.

  1. Meta Model API — Pricing and rate limits ↗

    Supports Wonder and Spark rates, search charges, pricing modes, the cached-input field and no separate long-context premium. Public page retrieved successfully.

  2. Meta Model API — Models ↗

    Supports Wonder and Full access, plan eligibility, mode distinctions and the fact that higher reasoning effort does not select Full. Public page retrieved successfully.

  3. Muse Code — Subscriptions ↗

    Supports usage multipliers, billing of additional API keys, Contributor settings, limits and plan management. Public page retrieved successfully.

  4. Meta — Muse Wonder model and plans ↗

    Lists Power at US$50/month and Maximum at US$200/month. This primary page was visible through the search/browser reader; a later direct retrieval returned a server error. We do not treat these figures as a verified checkout for every region.

  5. Meta AI Research — Muse Wonder release ↗

    Supports the 12 October 2026 release date and identifies the official artwork. Article and artwork retrieved successfully.

  6. Meta Model API — Overview ↗

    Official starting point for API access and model identifiers. Actual access is subject to current provider and account conditions.

  7. Meta Model API — Prompt caching ↗

    Further implementation reading. The price and cached-token description used here are also stated in the verified pricing page. We do not infer guaranteed cache behavior from this link.

Calculation method

Each estimate splits total input into cached and uncached tokens, applies the corresponding rate and adds output and search charges. Token charges are divided by one million; search costs $0.0025 per query. The per-request amount is multiplied by your chosen request count.

Calculations use unrounded values before formatting currency. Small per-request values show extra decimals. A fractional average search count is allowed, but the calculator does not send real tool calls. No API key or prompt is collected.

Examples are chosen workloads, not measured invoices. We did not run the model, test coding benchmarks or survey users for this guide. There are no fabricated testimonials or performance scores.

Asset and template provenance

  • The release artwork is the unchanged image from Meta AI Research’s announcement. Copyright belongs to Meta; it identifies the subject of this guide.
  • The favicon is the existing Lucide calculator icon. Its license is included.
  • We adapt Play Bootstrap by UIdeck on StartBuildingWeb. Its SaaS pricing-card, navigation and FAQ structure fits a developer pricing guide. The original Play stylesheet and pricing structure are retained; the project uses a smaller custom grid instead of loading the full Bootstrap stylesheet. Content, calculations and responsive refinements are project-specific.
  • The Play MIT license and Inter font license are preserved. No demo portraits or fake client logos are used.

No image-generation service was used. The pricing tables and calculator are HTML and arithmetic, not simulated screenshots of an account.

Verification limits

We have not verified consumer-app checkout prices for every region, a conversion from subscription allowances into API tokens, a Full API rate or a guaranteed cache hit rate. We make none of those claims.

When a price changes

Use the provider’s current price and checkout for spending decisions. The check date is an editorial snapshot, not continuous monitoring. Direct references on the API guide and the subscription guide let you check the current source.