Meta Releases Llama 4.5 With Native Multimodal Reasoning, Targets Enterprise Deployment Over Consumer Chat

Meta's Llama 4.5 release, which reached broad availability in late July 2026, marks a deliberate repositioning of the open-weight model line: rather than competing directly with Claude, GPT, and Gemini on chatbot experience, Meta is explicitly targeting the segment of enterprise buyers who cannot or will not send data to a third-party API for regulatory, data-sovereignty, or cost-at-scale reasons. The technical headline is native multimodal reasoning — the model processes image, text, and structured document inputs in a single forward pass rather than routing images through a separate vision encoder bolted onto a text model, an architecture change Meta says materially improves performance on tasks like extracting structured data from scanned documents or reasoning about charts and diagrams embedded in reports. The licensing terms remain the more consequential detail for enterprise adoption than the benchmark scores. Llama 4.5 continues under Meta's custom license, which permits commercial use but includes a clause requiring companies with more than 700 million monthly active users to obtain a separate license from Meta — a threshold that excludes all but a handful of the largest technology companies from any restriction, while still allowing Meta to describe the model as "open" for licensing and marketing purposes. For enterprises already running self-hosted inference infrastructure — banks, healthcare systems, and government contractors with data residency requirements that preclude API calls to Anthropic, OpenAI, or Google — Llama 4.5 is a genuine like-for-like option for the first time on multimodal document-heavy workloads, an area where prior open-weight releases lagged closed-API models by a wider margin than on pure text tasks.