Pandaily·· 13h agoAI score40
METASTONE launches multimodal inference service, reports 2.48x video speedup in self-tests
METASTONE Launches Multimodal Inference Service, Reports 2.48x Video Speedup in Self-Tests
Summary
Chinese computing infrastructure provider METASTONE has launched a multimodal model service for production workflows, combining video-generation acceleration, cache and memory management, token compression and a Context Engine that turns reference material into structured generation instructions. The company reports an average end-to-end speedup of 2.48 times over a dense-computation baseline with quality maintained, but these are vendor self-tests, not independently verified.
This site provides summaries and links to the originals only.
Source: Pandaily · pandaily.com