GPT-5.6 Luna (Global)

GPT-5.6 Luna is positioned within the broader GPT-5.6 family deployed on Amazon Bedrock as a lighter, speed-oriented endpoint alongside the Sol and Terra variants. The Bedrock deployment guide describes the three models as sharing a one-million-token that quick-info value foundation while differing in their model references, API, tool, routing, observability, and governance contracts across in-Region and cross-Region profiles. For Luna specifically, that positioning is framed around fast, high-volume workloads such as classification, routing, extraction, and summarization, where quick turnaround and predictable cost matter more than the deeper reasoning reserved for the Sol variant.

In practical terms, Luna inherits the family's shared multimodal and API surface, supporting text and image input with text output, and is reachable through Responses, Chat Completions, and Converse with prompt caching available. The combination of a very large that quick-info value window and a throughput-oriented design makes it a natural fit for production pipelines that need to process long documents or many short requests in parallel, rather than for the most demanding chain-of-thought problems. An independent model-selection write-up further highlights its cost-efficiency on agent-style coding tasks relative to some premium alternatives, reinforcing its role as the economical workhorse of the GPT-5.6 lineup on Bedrock.