OpenAI has officially launched its GPT-5.6 model family to the public after weeks of government review, marking one of the most closely watched releases in the company's history.

The rollout, which began on July 9, 2026, includes three distinct variants: Sol, a reasoning and coding-focused flagship; Terra, a balanced general-purpose model; and Luna, a cost-efficient tier designed for high-volume applications. Sol has drawn particular attention for achieving throughput of roughly 750 tokens per second on specialized hardware, making it one of the fastest frontier models available.

The release had been delayed by a U.S. government national security review that examined whether the models posed risks related to foreign access and misuse. OpenAI received broad approval to proceed, though the review process underscored the growing tension between rapid AI advancement and Washington's efforts to control frontier technology.

Enterprise customers and developers have quickly adopted the new family. OpenAI reported that Sol is already handling complex software engineering tasks, scientific research queries, and advanced data analysis for paying customers. Terra targets everyday ChatGPT users, while Luna is positioned to compete with lower-cost alternatives from Google and Chinese labs.

The launch arrives amid an intensifying race among AI labs. Meta released its Muse Spark 1.1 agentic model days earlier, and Google rolled out its Gemini 3.6 Flash line later in the month. Analysts say the crowded July release calendar reflects an industry-wide push to capture market share before potential regulatory restrictions tighten.

OpenAI also announced price cuts for the Luna variant later in July, signaling a willingness to sacrifice margin for scale. The company has not disclosed specific revenue figures tied to the GPT-5.6 launch, but internal projections described to investors suggest the family could account for a majority of API usage by year end.

Image source: i.ibb.co