← Back

Wire · technology

Scale Agentic Coding Workflows with Meta's Muse Spark Open on NVIDIA GB300 NVL72

Published

15 October 2026

Topic

technology

◆ Sectors

AI Frontier ModelsAI InfrastructureDeveloper Tools

◆ Geography

United States

◆ Source

Read at developer.nvidia.com →

◆ Verified

Fusion42 · 15 October 2026 · Fusion42 review

Meta has launched Muse Spark Open, an 816 billion parameter mixture-of-experts AI model optimized for agentic coding workflows, trained on NVIDIA's GB300 NVL72 Blackwell Ultra. It supports up to a 1 million token context window enabling complex project comprehension, includes multimodal perception across text, image, audio, and video, and runs efficiently with NVFP4 checkpoints to deliver high throughput and low cost per token.

This Wire brief sits within Fusion42's coverage of AI Frontier Models, AI Infrastructure and Developer Tools, and 21 sources have reported it between 5 Sep 2026 and 15 Oct 2026.

◆ ◆ The Wire takeaway

Agentic AI has advanced beyond single-shot coding to handle entire complex projects with multimodal inputs, letting you build tools that manage large codebases and multimedia context with unmatched scale and speed.

◆ Coverage

21 sources · first reported 5 Sep 2026 · latest 15 Oct 2026

◆ Related on Wire

◆ Topics

AI Frontier ModelsAI InfrastructureDeveloper Toolsagentic-aimultimodal-aicode-automationnvidia-gb300muse-spark