Pixtral Large (25.02) (US)

Pixtral Large (25.02) is a member of the Pixtral family of multimodal models, positioned as a larger-tier variant within that lineage. As a Bedrock-hosted endpoint, it is exposed to users through a managed cloud interface rather than as a standalone downloadable release. The "Large" designation within the Pixtral family suggests a capacity-oriented variant intended for more demanding multimodal workloads compared to smaller siblings, though no official parameter count, architecture diagram, or scaling disclosure was available in the supplied evidence pack for this exact Bedrock-hosted build.

In practical terms, the model's defining characteristic is its ability to accept both text and image inputs in a single workflow, making it suitable for tasks such as document interpretation, visual question answering, and image-grounded analysis. The combination of a long context window with vision input support points to use cases involving lengthy documents that include embedded figures, charts, or screenshots. Without benchmark disclosures or architecture specifics in the supplied sources, practitioners evaluating this variant should treat it as a general-purpose multimodal option suited to image-plus-text reasoning pipelines where Bedrock integration is already in place.