PulseAugur
EN
LIVE 23:08:57

Meta releases Muse Spark Open, an 816B multimodal coding agent model

By PulseAugur Editorial · [1 sources] ·

Meta AI Research has released Muse Spark Open, an 816B-parameter mixture-of-experts model designed for coding and agentic tasks. This open-weight model supports multimodal inputs including text, image, audio, and video, and boasts a 1 million token context window. Developers have noted that the model's reasoning capabilities are always active and consume tokens from the same budget as the output, requiring careful management of the `max_tokens` parameter. Additionally, the KV cache size is significantly smaller than what might be expected for a model of its parameter count, due to a hybrid attention mechanism that limits the full context to a subset of layers. AI

IMPACT This release provides a powerful open-weight multimodal model for coding and agentic tasks, potentially accelerating development in these areas.

RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Meta releases Muse Spark Open, an 816B multimodal coding agent model

How we ranked this

Signal score
48 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Significant
Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
model release, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Yong Yu ·

    Muse Spark Open Is Out: Reasoning You Can't Turn Off, KV-Cache Math at 128K vs 1M, and Hosted vs Mac vs GPU

    <blockquote> <p><strong>Attributed compile + two small offline scripts</strong><br /><br /> <strong>Primary sources (all 2026-10-15):</strong> Meta AI Research, <a href="https://research.meta.ai/blog/muse-spark-open" rel="noopener noreferrer">Introducing Muse Spark Open</a>; Hugg…

RELATED ENTITIES

RELATED TOPICS