Muse Spark 1.3 is Meta’s frontier proprietary reasoning model, released on September 2, 2026 and available now through Meta’s first-party Model API and its Muse Code coding agent. It keeps the 1M-token context window of Muse Spark 1.2, accepts text, image and video input, and is priced unchanged at $1.25 per million input tokens and $4.25 per million output tokens, with cached input at $0.15.
The generally available “xhigh” variant scores 61 on the Artificial Analysis Intelligence Index, up 4 points from Muse Spark 1.2 and tied with GPT-5.6 Sol (max) and Grok 4.6 (high); a “max” variant scoring 62 is in limited preview for Meta partners. The gains are concentrated in agentic work and science, including a 12-point jump on Tau3-Bench Banking (35% to 47%) and a higher GDPval-AA v2 Elo of 1,709, and at $0.55 per Intelligence Index task it is the cheapest model scoring 59 or above.
Source: Muse Spark 1.3: Meta reaches the frontier — Artificial Analysis