Meta Releases Llama 4 Scout with 10 Million Token Context for Ent...

Meta released Llama 4 Scout, the enterprise-grade variant of its open-weight Llama 4 family, featuring a landmark 10-million-token context window that can ingest entire corporate document repositories, multi-year correspondence archives, or complete software codebases in a single inference call. Scout is optimised for retrieval-augmented generation tasks and demonstrated superior performance to closed-source competitors on enterprise reasoning and long-document question answering benchmarks. Unlike consumer Llama variants, Scout is available under a commercial licence and includes Meta safety and content moderation layers. The model can run on eight H100 GPUs, making it affordable for mid-size enterprises without hyperscaler infrastructure. Wipro, Infosys, and HCL Technologies have each announced Llama 4 Scout-based enterprise AI products for Indian and global markets.