Meta releases Muse Spark Open weights

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

Meta released the weights of Muse Spark Open — a natively multimodal model for coding and agents with 816 billion total and 42 billion active parameters and a context of 1 million tokens, on Hugging Face under the Muse Model License

Why it matters

Meta claims it is the most capable US open-weights model to date. It adds safety post-training for standalone use on top of the same base model as Muse Spark 1.3, and ships with SGLang and vLLM optimizations and Helion kernels (1.03–3.95 times faster on GB300). Muse Spark, which began as a private preview in April, reached open weights within half a year

Source — Primary sources first; secondary reports are labelled “Press”

Officialpytorch.orghttps://pytorch.org/blog/muse-spark-open/

Added in Week 42, 2026 · See that week

On the map — ⑨ Open-Weight Lineage line

  1. Previous stopK-EXAONE 2.0
  2. This recordMuse Spark Open
  3. Next stopMistral AI releases the weights of Mistral Large 4 (Le Chonk)
See it on the map

In the model family tree — Llama family · 12 versions

  1. Earlier versionMuse Spark 1.3
  2. This recordMuse Spark Open
  3. Later versionThe latest version in this branch
See the whole Llama family

Same week

Anthropic commits 5% of its equity to the Anthropic Public Benefit FundOfficialReflection launches Beam, an open-weight modelOfficialNVIDIA previews a 2.8 trillion token corpus for 22 Indian languagesOfficialNVIDIA launches RTX Spark laptopsOfficial

Nearby on the same axis

Mistral AI releases Timestral, a time-series forecasting modelOfficialReflection launches Beam, an open-weight modelOfficialAllen Institute for AI releases AstaBrief 8BOfficialMistral AI releases the weights of Mistral Large 4 (Le Chonk)Official