Meta's Muse Spark 1.3 on one card: the 11 tests decoded, the win-loss vs Claude and GPT, the price math, and the 3 games

Meta released Muse Spark 1.3 on September 2, 2026, one day after Claude Fable 5.1. This field card is the launch in plain English: what the model is and where it runs (Muse Code and the Meta Model API), all eleven benchmarks from Meta's evaluation report translated into one line each with every score (Muse Spark 1.3 max and xhigh, Muse Spark 1.2, GPT-5.6 Sol, Claude Opus 5), the honest win-loss tally (all three coding tests won against Opus 5, every office-style test lost by a point or two; 8 wins, 1 tie, 2 losses against GPT-5.6 Sol), the price math (standard $1.25 in / $4.25 out vs Opus 5 at $5 / $25, and the Contributor tier at $0.10 / $0.20 with its training-data condition), the two catches (max reasoning not out yet; Contributor terms and country limits), the one-line install for Muse Code, and links to the three playable demos Meta says the model coded: Psychic Storm, Stimm Tuner, and Startup City 3D. Independent explainer from Hyperautomation Labs; every number is sourced to Meta's own report and pricing page.

By submitting, you'll be subscribed to the Hyperautomation AI Report newsletter — one practical AI email per week. Unsubscribe anytime.

Free. No spam. Unsubscribe anytime.