Blog
Research
Stories, research, and updates from the Meshy community
Research
Meshy T2 turns one image into a clean, compact, multi-part 3D mesh in a median 6 seconds, using flow matching instead of iso-surface extraction.
Meshy T2: Native 3D Mesh Generation in 6 Seconds
Rendong Liang· 2 days ago
Research
Inside Muse, Meshy's internal model evaluation platform: how a boring stack of Slack, Linear, GitHub, and Claude Code makes AI-written code trustworthy.
Plain Tools, Self-Driving Development: How Muse Builds Itself
Andrew· 11 days ago
ResearchFaithful Contouring: No SDF, No Watertight, 1024³ Voxelization
Robin· 2 months ago
Research
How Meshy built Overmind — an open-source library that cuts ML model loading from 15s to 0.2s using zero-copy shared memory, requiring no changes to inference code.
Overmind: Cut ML Model Loading from 15s to 0.2s
Bin Wang· 5 months ago