20 Dollars or 100 Dollars a Month: Pricing the Muse Consumer Subscription

When Muse launched on September 8, 2026, the pitch was deliberately simple: free for most features, with Meta planning to make money on a small transaction fee rather than a hard paywall, as covered in Free By Default: Why Meta Priced Muse the Way It Did. A month later, with Muse Wonder's October 12 launch, actual subscription numbers have entered the picture for the first time — and they arrive at exactly the moment where it's easiest to conflate two genuinely different price lists: what a developer pays to use the Wonder model through Meta's API, and what a consumer pays for the Muse app itself.

Two separate prices, two separate products

The consumer-facing Muse app subscription is reported, per CNBC's coverage of the October 12 announcement, to run $20 or $100 a month, depending on tier. That's a subscription for the app experience — the thing an ordinary user downloads and talks to — and it's separate and distinct from what Meta charges developers to call the Wonder model directly through the Meta Model API, which is priced at $4 per million input tokens and $20 per million output tokens. Those two numbers happen to share a "$20" in them, which is exactly the kind of coincidence that makes it easy to accidentally quote the wrong one in the wrong context. The API price is a cost of computation, scaled to how much text goes in and comes out; the consumer subscription is a flat monthly fee for using the product, the same category of price as a Netflix or Spotify tier.

Worth noting as a caveat: the $20/$100 consumer figures and their precise tier boundaries come from one source in the reporting gathered for this piece, and an independent re-fetch of that reporting wasn't possible during this article's research — treat the exact framing of what each tier includes as less certain than the headline numbers themselves, pending a direct look at Meta's own pricing page.

What each tier is actually buying

Here's where the real ambiguity sits: Meta hasn't published a clear public breakdown of what separates the lower and higher consumer tiers. What's known is the model access they presumably gate — Wonder Full, the larger of the two new models, is explicitly restricted to Muse Code subscribers and isn't available through the API at all, while the smaller Wonder model is. If the consumer app's two subscription tiers map onto model access the way that restriction suggests, the higher tier likely corresponds to Wonder Full access and the lower one to standard Wonder or continued Muse Spark usage — but that's an inference from the Wonder Full/API restriction, not a confirmed mapping Meta has published, and it should be labeled as such rather than stated as settled fact.

This is also the first real test of the "free for most features" promise from launch. Meta's original framing was that most users would never hit a paywall, with paid tiers reserved for heavier usage. A $20-or-$100 subscription structure is consistent with that framing if the free tier remains genuinely capable for typical use and the paid tiers are aimed at power users who want Wonder Full specifically — but it's also exactly the kind of structure that could, in practice, migrate toward gating increasingly capable models behind payment as Meta's frontier-model lineup grows. Which of those two outcomes this turns into is a question about Meta's future product decisions, not something the current pricing announcement answers on its own.

Why the API price matters for a different reason

The $4-input/$20-output API pricing is worth its own scrutiny for a different reason: it's being explicitly compared, in Meta's own positioning per reporting on the launch, to GPT-6 Sol and Opus 5.5 — meaning Meta is pricing Wonder to compete directly with the cheaper, faster tier of frontier models from OpenAI and Anthropic, not against their most expensive flagship offerings. That's a meaningful competitive signal distinct from the consumer subscription question entirely: it tells you where Meta thinks Wonder fits in the market for developers building their own products on top of someone else's model, which is a different audience and a different decision than someone deciding whether to pay $20 a month for the Muse app.

The cleanest way to keep these straight: the API price is what it costs to rent Wonder's intelligence by the token if you're building something; the subscription price is what it costs to use Meta's own finished product. Two different buyers, two different price structures, and conflating them is the easiest factual error this specific pricing story invites.

Where this leaves the "free for most features" promise

The honest reading of this moment is that Meta hasn't broken its original promise, but it has started drawing the lines that promise always implied would eventually exist. "Free for most features" was never a claim that everything would stay free forever — it was a claim about where the line would sit, and a $20-or-$100 subscription structure arriving alongside, not instead of, continued free access is consistent with that original framing rather than a reversal of it. What would actually constitute a reversal is if the free tier's capability meaningfully degrades over time as Meta pushes more of its best models behind the paid tiers — something there's no evidence of yet, but which is exactly the kind of slow-creep risk worth watching for in a product built around "free now, monetized later" rather than a hard paywall from day one. The transaction-fee revenue model Meta originally described, separate from subscriptions entirely, also hasn't been quantified anywhere in the reporting gathered for this piece — meaning there are now, in principle, three different Muse revenue streams (transaction fees, consumer subscriptions, and API usage fees) with only one of them, the API pricing, actually specified down to a real number.

What's still missing

What hasn't been published, as of this writing, is a side-by-side feature comparison of the two consumer tiers, a usage-limit disclosure for either one, or confirmation of exactly when Wonder and Wonder Full roll out to the consumer app versus remaining Muse Code/API-only, since Meta's own language described that consumer rollout as happening "soon" rather than immediately. Until that detail lands, the honest state of Muse's pricing is: two real numbers, a plausible but unconfirmed guess at what separates them, and a developer-facing API price that's doing more to signal Meta's competitive position than to tell an ordinary user anything about what they'd actually be paying for.

Part 44 of 46 in the Meta Muse series. Previously: The Personal-Messages Complaint: What Is Confirmed and What Isnt. Next: Muse vs Dots vs Grok Bot, One Month Later: What Actually Shipped. Browse more builder insights or explore AI skills at aiskill.market.