Meta's Muse Already Lost the App Store to ChatGPT — Now It's Targeting 33 Million Small Businesses

Consumer Muse launched on September 8, 2026 powered by Muse Spark 1.3, the first Meta Superintelligence Labs (MSL) release to publish frontier-tier reasoning scores. Meta's own blog reports Spark 1.3 in \"Contemplating mode\" hit 58% on Humanity's Last Exam and 38% on FrontierScience Research — competitive with the deepest reasoning modes on GPT and Gemini. The model carries a 1M-token context window and runs the same $1.25-per-million input / $4.25-per-million output API price as Spark 1.2, with a $0.10 / $0.20 contributor tier that undercuts OpenAI and Anthropic on cost-to-serve.

That last number is the wedge. Microsoft Copilot has been on the market three years; even after passing 30 million paid seats it covers under 5% of the installed base, and adoption skews to enterprise IT departments — not to the dry cleaner, the yoga studio or the Etsy seller. SMBs are exactly the long tail that Copilot has not landed in.

Meta's pricing matrix is the most aggressive of the group: a fully featured free tier with usage limits, a $20 Power plan that beats every per-seat competitor's entry price, and a $100 Maximum plan that lands exactly where ChatGPT Pro and Microsoft Copilot's premium SKUs sit. The $20 Power and $100 Maximum token allowances —500M and 3B tokens weekly, respectively — are published by Meta, not estimated.

Fifteen third-party integrations launched with Muse for Small Business on day one — Shopify, Intuit QuickBooks, Stripe, Slack, Asana, Box, Canva, Figma, Granola, HighLevel, Klaviyo, Lovable, Notion, Zoom and Dropbox. Three Meta-owned data sources sit alongside them: Instagram professional account analytics, Facebook Pages and Meta ad accounts.

That last connector is the structurally differentiating one. When a dry cleaner asks Muse to \"find new customers,\" the agent can pull real performance data from the owner's existing Meta ad account and draft new campaigns that actually buy against the same targeting that already worked. No competitor can touch that without first negotiating an integration that Meta owns.

On the night of September 20, 2026, Amazon cut off Muse from completing purchases on Amazon.com. The retailer cited data security and consent: Meta never asked, the agent does not identify itself when browsing, and the bot can capture credentials or reach order history if prompted. Users see a popup that an \"unauthorized AI agent violates Amazon's Conditions of Use.\" Amazon had separately launched its own Buy for Me AI in the same window.

Forbes framed the episode bluntly: Amazon has $68 billion in annual ad revenue on the line — the same storefront Muse was driving traffic toward is the one Amazon spends that money to fill. Cutting the agent off protects both sides of the ad business.

Free Muse consumers will eventually hit usage caps and upgrade. SMB owners paying for Power and Maximum tiers generate token-usage revenue that compounds. That monetization path is what justifies the spending Meta Platforms is doing on the front end.

Meta's Q2 2026 financials show the bill: revenue of $60.801B (+28% YoY) supported a Family-of-Apps operating margin that funded a $31.08B capex quarter, leaving free cash flow down ~91% YoY at roughly $784M. The full-year 2026 capex guide sits at $130–145B (narrowed from $125–145B at Q1), with total expenses of $162–169B. The stock dropped about 10% the day Q2 reported, even with ad revenue up 27% Y/Y to $59.363B. Reality Labs lost another $4.619B that quarter.