Muse Compute (Credits)

Muse Compute is the single balance that powers the Muse agent, Vital AI, and M4L requests for Free and Pro users. This page explains how you earn it, how it is spent, and when you are not charged.

Earning Muse Compute

Free (daily). Free users receive 5 Muse Compute per day. A scheduled job tops up your balance when it is below 5; unused credits do not roll over to the next day.

Pro (monthly). Pro subscribers get 200 Muse Compute each month, granted when your subscription invoice is paid (e.g. at the start of each billing cycle or when you first subscribe). Trialing subscriptions do not receive the grant until the first paid invoice.

Boost packs. One-time top-ups of 50, 110, or 250 Muse Compute. These are added to your existing balance and do not replace it. You can buy them from the pricing and account pages.

Promo codes. Redeemable codes that grant a fixed amount of Muse Compute. Redemption is available in the Muse agent Settings and in upsell modals; each code can only be redeemed once per user.

M4L one-time purchase. The one-time Muse Max for Live purchase is an entitlement only. It does not grant Muse Compute; it unlocks M4L access. Your Muse Compute balance is not used when you make M4L requests if you acquired the device through a one time purchase.

Spending compute

Agent chat. Compute is deducted after each completed agent run. There is no upfront reserve: we charge only when the run finishes. The cost is based on token usage and the model you chose.

Cost formula.

Credits are computed using the folling formula:

credits=110⌈10⋅max⁡(0.5,  effectiveTokenstokensPerUnit⋅modelMultiplier)⌉\text{credits} = \frac{1}{10}\left\lceil 10 \cdot \max\left(0.5,\; \frac{\text{effectiveTokens}}{\text{tokensPerUnit}} \cdot \text{modelMultiplier}\right)\right\rceil

Where

effectiveTokens=tokensIn+outputWeight×tokensOut\text{effectiveTokens} = \text{tokensIn} + \text{outputWeight} \times \text{tokensOut}

And

tokenPerUnit∈[10,000, 50,000]\text{tokenPerUnit} \in [10{,}000,\, 50{,}000]
outputWeight∈[4, 8]\text{outputWeight} \in [4,\, 8]

Are constants adjusted depending on usage rates.

What this means for you: The normal request is 1–2 credits/run, and a complex request with a premium model is in the ballpark of 25 credits/run.

Model tiers. Models have different multipliers (e.g. 0.5× for fast, 1.5× for standard, 2.2×–3.2× for premium). Premium models (e.g. GPT‑5.2 High, Claude Sonnet 4.5, Claude Opus 4.6) are Pro-only. The model selector in the app shows cost tiers such as Fast, Standard, and Premium.

Other products. Vital AI, sample generation, MIDI generation, and M4L requests also draw from the same Muse Compute balance. Each has its own cost logic, but they all debit the shared balance.

When you're not charged

Non-final finish reasons. Runs that finish with tool-calls or error are not charged. Normal completions (e.g. stop, length) are charged.

Storage and display

Balances are stored in milli-credits internally (1 credit = 1,000 milli-credits) for precision. The API and UI convert to credits for display by dividing by 1,000, so the numbers you see (e.g. in the agent, on the account page) are in credits.

Low balance

A low-balance warning is shown when your balance is at or below a configured threshold. If your balance is too low to cover the minimum cost of a run (0.1 credit), the request will not start and you may see an insufficient-compute message (HTTP 402) or an upsell modal. You can add Muse Compute via Pro or Boost packs, or by redeeming promo codes in the agent (Settings or upsell modals).

View pricing →