gpt-5.6-luna

Vision capability plus large context window supports OCR-style document tasks, though not marketed as a top-tier vision reasoning model.

1.05M-token context window allows ingesting very long documents, but as the fastest/lowest-cost tier it trades off some depth versus higher GPT-5 variants.

Low-cost, high-throughput profile suits large-scale entity extraction tasks over the model's 1.05M-token context window.

Positioned for lightweight agentic tasks — adequate for straightforward function-calling apps where speed and cost outweigh the need for deep reasoning.

Vision capability is present, and the low-cost tier is well suited to high-volume image Q&A workloads rather than complex visual reasoning.

Vision-enabled and optimized for low-cost, high-volume tasks — a reasonable fit for lightweight screenshot interpretation workflows.

Positioned for lightweight agentic tasks, not complex coding agents; family's higher-tier variants (e.g. gpt-5.4 at 71.1 coding index) are better fits for agentic coding.

Extended-thinking is present, but the model is explicitly the fastest/lowest-cost tier built for lightweight tasks, not deep multi-step planning — heavier GPT-5 variants are better suited.