OpenAI released GPT-5.6 Sol on July 9, raising the bar for what a large language model can do. Sol ranks as the company’s flagship AI, designed to handle coding, scientific reasoning, and complex multi-step planning with fewer missteps than competing systems.

The model supports a 1.05 million token context window—meaning it can read and remember vast amounts of text in a single conversation. That capability matters for researchers, engineers, and analysts who work with large documents or codebases.

What Sol Does Better

GPT-5.6 Sol claims state-of-the-art results in coding, knowledge work, cybersecurity, and science. OpenAI didn’t just increase model size; they focused on reliability and reasoning. The model handles long-horizon planning, meaning it can break down complex tasks into sequences of steps and execute them.

For agentic workflows—where AI systems make decisions and use tools autonomously—Sol improves how the model reasons through problems. Fewer hallucinations. Better fact-checking. Cleaner code generation.

The Lineup

Sol is the top-tier option. Terra sits in the middle for everyday work and balanced pricing. Luna is the most cost-efficient for those who prioritize speed over maximum capability.

OpenAI also released Sol on Cerebras hardware, reaching 750 tokens per second—a speed that lets customers stream responses in real time without waiting.

Pricing

Standard short-context pricing sits at $5 per million input tokens, $30 per million output tokens, with caching options for repeated inputs. Cost-per-token is lower than previous flagship models, making frontier AI accessible to more teams.

GPT-5.6 Sol represents incremental progress on the frontier. It’s faster, more reliable, and cheaper than what came before.

FYI (keeping you in the loop)

What’s a token context window?

It’s how much text the model can read at once. A 1.05 million token window means Sol can analyze roughly 700,000 words in a single conversation without forgetting earlier content.