Polylog
Sign inSubscribe
← The Polylog AI Intelligence Brief

Morning Edition · Sunday, July 19, 2026Published at 1:30 AM EDT · New York

Meta Opens Its First Paid Model API With Muse Spark 1.1

The million-token agentic model prices at 1.25 and 4.25 dollars per million input and output tokens, putting Meta directly into the coding-agent market against Anthropic and OpenAI.

Save
Meta Opens Its First Paid Model API With Muse Spark 1.1

Meta on July 9 introduced Muse Spark 1.1 alongside a public preview of its new Meta Model API, the first time outside developers can pay to build on a Meta frontier model. The model, from Meta Superintelligence Labs, is a multimodal reasoning system with a one-million-token context window, tuned for tool and computer use, coding, and multimodal understanding.

The move is a strategy shift as much as a product. Meta established its reputation on downloadable Llama weights. A metered API at 1.25 dollars per million input tokens and 4.25 per million output tokens, with 20 dollars in starter credits, puts Meta on the same commercial basis as the closed labs it once positioned itself against. CNBC framed the launch as Meta pursuing Anthropic and OpenAI in coding specifically.

On Chamath Palihapitiya's price scale, Meta's list price sits well below the frontier US labs and well above the lowest Chinese prices, which positions Muse Spark as a mid-price agentic option rather than a premium reasoning model. The coding claims come from the vendor and await independent benchmark reproduction.

What this means

The channel is distribution and revenue. Meta is following the same move away from open weights toward metered APIs that the other Western labs made, pursuing the coding-agent budget where usage and revenue concentrate. The exposed party is the open-weight developer ecosystem that relied on Meta as its main supplier and is now pushed toward Chinese open weights or paid US APIs.

What to watch

  • Independent coding benchmark results for Muse Spark 1.1 against Claude and GPT-class models, which would show whether the mid-tier price reflects mid-tier capability.
  • Whether Meta keeps shipping downloadable weights in parallel, since a full shift to paid-only would confirm the industry-wide move to close the weights.

Observations to monitor, not financial advice.

2 sources

Synthesized from: Meta AI · CNBC

Part of a tracked trend

US Labs Close Weights to Monetize

Western frontier labs increasingly retreat from open weights toward closed, metered APIs to capture revenue, ceding the downloadable open-weight tier to Chinese suppliers and splitting the market by license as much as by capability.

More from this edition