OPEN LINK ↗
// 1h agoMODEL RELEASE
Muse Spark Open Releases 816B Weights
Meta released Muse Spark Open, an open-weight multimodal MoE model with 42B active parameters, 816B total parameters, and a 1M-token context window. It targets coding and agentic workflows, with downloads, hosted APIs, and broad inference-engine support available today.
// ANALYSIS
The real story is distribution: Meta is turning a frontier-scale model into something developers can self-host, customize, and run across a growing hardware ecosystem.
- –Sparse routing keeps per-token compute closer to a 42B model, but serving still requires loading the full 816B checkpoint
- –One-million-token context makes it a strong candidate for repository-scale coding and long-running agent workflows
- –Open weights enable private deployment and fine-tuning through PyTorch, while hosted providers reduce the infrastructure barrier
- –Multimodal inputs plus support for SGLang, vLLM, TensorRT-LLM, MLX, llama.cpp, and Unsloth broaden practical deployment options
// TAGS
muse-spark-openllmopen-weightsmoelong-contextai-codingcoding-agentmultimodal
DISCOVERED
1h ago
2026-10-15
PUBLISHED
2h ago
2026-10-15
RELEVANCE
10/ 10
AUTHOR
MetaforDevs