Meta launches LlamaX-Seq for long-context prototyping and multimodal knowledge retrieval

AI · 5 min read

Meta launches LlamaX-Seq for long-context prototyping and multimodal knowledge retrieval

LlamaX-Seq expands upon Meta’s Llama family with improved long-sequence handling and tighter image grounding, allowing the model to ingest entire design systems, component libraries, and asset folders. Designers can query the model about previous design decisions without reuploading assets.

The model supports session memory that persists across authenticated team sessions, which helps keep style guides, token sets, and revision rationale available to design-assisted prompts. Meta says it has improved grounding to reduce hallucinations when the model generates design suggestions tied to a specific system.

Meta packaged a developer SDK with connectors to Figma, Notion, and internal CMSs to make it easier for teams to use LlamaX-Seq as a collaborative design assistant. Privacy controls and on-prem options are offered for enterprise customers who need stricter data governance.