Muse Wonder pricing: API costs, subscriptions & calculator

Subscription plans and model access

Official documentation· Last verified 2026-10-14

Usage limits apply. These are not API credits; check taxes, region and eligibility with the provider. Meta ↗

Muse Wonder pricing: API billing and subscriptions

Editorial synthesis · Cited sources· Last verified 2026-10-14

The API charges for metered usage. A Muse Code subscription pays for access to the coding product within its usage limits. A monthly subscription price cannot be treated as a per-token quote, and this token calculator does not estimate your Muse Code allowance.

Wonder is listed on every Muse Code plan. Full is listed on Power Usage ($50/month) and Maximum Usage ($200/month). Those subscriptions do not turn Full into a public API model.

Read your usage before estimating your bill

Editorial synthesis · Cited sources· Last verified 2026-10-14

Use total input tokens for the period, then enter the subset actually served from cache. The calculator subtracts cached input from total input before charging the ordinary input rate, so the same token is never counted twice.

Enter all billed output tokens, including any billed reasoning tokens reported by your provider, and the actual number of search queries. A tool can be called repeatedly during one agent run; one user message is not necessarily one billable request.

What this estimate leaves out

Editorial synthesis · Cited sources· Last verified 2026-10-14

This is a list-price estimate in USD, before taxes, third-party platform fees, credits, retries and any other separately billed tools. Cache eligibility is determined by the provider; entering a cache count here does not make your requests cacheable.

At these list prices, 1 million ordinary input tokens plus 100,000 output tokens costs $6 with no searches. A longer generated answer can cost more than a large prompt, which makes output budgets worth measuring separately.

Muse Wonder pricing formula: reproduce the calculator

Editorial synthesis · Cited sources· Last verified 2026-10-14

The token estimate is ((total input − cached input) × 4 + cached input × 0.20 + output × 20) ÷ 1,000,000. Add search queries × 2.50 ÷ 1,000 for the search estimate. All four quantities describe the same billing period. Enter raw token counts rather than millions: one million tokens is entered as 1000000. The calculator does not convert words, characters, pages or minutes into tokens, because those conversions depend on the actual request and the provider’s accounting.

For a reproducible example, enter 1,000,000 total input tokens, 200,000 cached input tokens, 100,000 output tokens and 1,000 searches. Ordinary input is 800,000 tokens and costs $3.20. Cached input costs $0.04. Output costs $2.00 and search costs $2.50. The combined estimate is $7.74. Removing the searches changes the estimate to $5.24. These are fixed-quantity examples you can verify in the calculator, rather than a promise about how much an application will spend.

Compare three Muse Wonder API cost scenarios

Editorial synthesis · Cited sources· Last verified 2026-10-14

A short-response workload with 1,000 calls, each using 2,000 uncached input tokens and 500 output tokens, totals 2 million input and 500,000 output tokens. At the listed rates, that is $8 for input and $10 for output, or $18 before tools and other charges. If calls vary in length, sum the recorded usage instead of multiplying one unusually short or long example by the whole request count.

A repeated-context workload with 2 million total input tokens, of which 1.5 million are actually cached, and 500,000 output tokens costs $2 for ordinary input, $0.30 for cached input and $10 for output: $12.30. This example shows why a cache-hit assumption materially changes the forecast. Measure the billed cached portion from responses; do not classify an entire conversation as cached simply because it repeats a system prompt.

A research workload using the same token quantities as that repeated-context example and 3,000 billed searches adds $7.50, giving $19.80. Search queries are counted independently from user messages and completed answers. An agent can search several times before answering. Record searches separately when evaluating tools so that a lower token bill does not hide a higher tool bill.

Build a monthly budget from observed usage

Editorial synthesis · Cited sources· Last verified 2026-10-14

Start with a small sample of representative tasks: a short question, a typical document, a long conversation and a tool-heavy request. Save the model identifier, date, total input, cached input, output, search count and whether the task was completed successfully. Keep these records in your own application logs with appropriate access controls. This site does not receive your usage records or connect to your billing dashboard.

Estimate each task category separately, then multiply its average recorded cost by its expected monthly volume. Keep a separate allowance for experiments, repeated attempts and longer-than-usual outputs. For example, 10,000 tasks at an observed $0.018 per task yield a $180 baseline. A planning allowance of 20% would produce a $216 budget; that allowance is your assumption, not a provider surcharge or an amount this calculator automatically adds.

Track cost per successful task alongside cost per request. If two approaches have identical token rates but one needs several attempts, the effective cost of the completed result can differ. Compare comparable tasks with similar answer-length limits and tool access. A benchmark score alone cannot tell you the monthly cost of your own application.

Meta direct and OpenRouter billing: check the whole invoice

Editorial synthesis · Cited sources· Last verified 2026-10-14

The direct Meta endpoint and OpenRouter use separate accounts, credentials and billing records. Matching advertised token rates do not establish identical final invoices. Before choosing a route, check credit-purchase terms, any platform fees, tax handling, supported tools and your account’s spending controls on that provider. A balance funded with one provider should not be assumed to pay for requests sent to the other.

When comparing a test, keep the same model version, prompt, output limit and tool configuration. Save each provider’s reported usage and the corresponding dashboard charge. If the totals disagree with this estimate, first check whether cached input was counted as a subset, whether all billed output was included and whether the test made additional requests. The calculation here is transparent arithmetic; the provider’s usage record remains the evidence for an actual charge.

Choose API usage or a Muse Code subscription

Editorial synthesis · Cited sources· Last verified 2026-10-14

Use an API estimate when you are budgeting requests from your own software. Use the Muse Code subscription table when you are deciding which coding-product access to purchase. A $50 monthly subscription and a $50 API forecast describe different services and cannot establish a token-equivalent break-even point. Wonder Full access also changes the choice: a published subscription entitlement does not imply a public Full API endpoint.

Before subscribing, check the current plan’s eligible models, usage limits, reset period, data mode, renewal conditions and cancellation options in the official product. Before building with an API, check model availability, rate limits and your account budget. The tables on this page are a dated snapshot with linked sources, and the calculator is an estimate. Confirm the current offer with the provider before paying or committing a production workload.