DigestAI news desk

Cut through the AI noise.

Generative AI & Models7 min read

Meta shifts to new flagship AI model Muse Spark

Meta has announced Muse Spark, a closed‑weight multimodal reasoning model, as its new flagship AI. The first public release is version 1.3, which can analyze text and images, use tools, and run sub‑agents. The model is available through an API and is priced at $1.25 per million input tokens and $4.25 per million output tokens, with a discounted Contributor tier.

1 source

Key points

  • Meta’s flagship model Muse Spark is a closed‑weight multimodal reasoning model released as version 1.3.
  • Muse Code offers a $5/month tier for a terminal‑based coding agent using Muse Spark, priced lower than Claude.

Muse Code is a terminal‑based coding agent that uses Muse Spark and offers a $5/month tier, cheaper than Claude’s $20/month plan. The personal‑assistant app Muse runs on a virtual Linux VM and can browse the web, make purchases, and interact with third‑party apps, but has raised security concerns. Meta also announced Muse Charm, a voice‑controlled physical AI device.

The shift follows Llama 4’s underperformance against Chinese open models and Meta’s reorganization into Meta Superintelligence Labs. The company is moving away from open‑source models toward proprietary offerings, a change that may alter the competitive landscape for developers and enterprises.

Model pages: Muse → · Muse Spark 1.3 →

The story so far

4 episodes →
  1. Meta shifts to new flagship AI model Muse Sparkthis story
Full story from Zapier Blog · by Harry GuinnessOpen source ↗

Meta AI Muse Spark: A guide to Meta's AI models

Zapier Blog · 29 September 2026

Loading the full article…

This text was published by Zapier Blog and written by Harry Guinness. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗

Topics · follow one to build your own front page

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.

Comments

via GitHub Discussions

More in Generative AI & Models

All →

Related stories