Meta Releases Muse Spark 1.3 with Improved Coding and Visual Capabilities
Meta has updated its Muse Spark model family to version 1.3, delivering incremental enhancements in code generation and visual tasks at a low cost. The release improves upon version 1.2 by demonstrating better SVG visual rendering and stronger coding benchmark scores. The update provides software developers with an affordable, reliable option for AI-assisted coding and multimodal workflows without needing costly frontier models. It highlights Meta's push to deliver competitive utility and performance in budget-friendly models. Users report noticeable improvements over version 1.2 in SVG generation quality, such as better anatomical and structural accuracy. Benchmark discussions highlight strong achievements, including top performance on DeepSWE benchmarks at a fraction of the cost of competing models.
## BACKGROUND
Developed by Meta Superintelligence Labs (MSL), Muse Spark is Meta's multimodal reasoning model designed for coding, tool use, and visual reasoning. Previous versions, such as Muse Spark 1.2, offered broad context windows and optimized coding workflows, serving as budget-friendly tools for software engineering tasks.