OpenAI is expanding its GPT-6 lineup with two new models, GPT-6 Sol and GPT-6 Luna, bringing some of the capabilities introduced with its flagship GPT-6 Astra to lower-cost and more accessible systems.
The new models arrive just weeks after OpenAI introduced GPT-6 Astra, which the company positioned as its most capable AI model yet. Rather than competing directly with Astra at the top end of the market, Sol and Luna are designed to make advanced AI capabilities more practical for everyday and high-volume workloads.
OpenAI describes GPT-6 Sol as a model aimed at more demanding professional tasks, particularly coding and other complex workflows. Luna, meanwhile, is positioned as the more economical option for repetitive tasks with clearly defined objectives, such as summarizing documents, extracting information and answering straightforward questions.
One of the biggest changes comes in pricing. OpenAI says API access to both new models will cost 50% less than the promotional pricing of their GPT-5.6 predecessors. GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens, while GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens.
OpenAI attributes the lower prices to improvements in caching and inference efficiency. Instead of simply increasing the capabilities of its models, the company is increasingly focused on making those capabilities cheaper to operate, a trend that is becoming particularly important as businesses deploy AI agents at larger scale.
The company is also claiming significant improvements in reliability. According to OpenAI, GPT-6 Sol makes roughly half as many factual mistakes as its predecessor in the company’s internal evaluation based on real-world conversations where users had identified incorrect answers. OpenAI says this brings Sol closer to Astra-level reliability while requiring significantly less computational resources.
The new models are also intended to improve coding performance, reasoning and alignment while retaining the efficiency needed for high-volume applications. OpenAI says Sol and Luna were developed using methods similar to those behind GPT-6 Astra, allowing the company to transfer some of the advances from its flagship model into smaller systems.
The launch also highlights the increasingly intense competition between OpenAI and Anthropic. OpenAI is positioning Sol and Luna against Anthropic’s latest models, claiming advantages across several tasks. The timing was particularly notable, as Anthropic launched a new version of its own flagship model, Opus 5.5, only around 90 minutes before OpenAI announced its new GPT-6 models.
That back-to-back timing reflects a broader shift in the AI race. The competition is no longer focused solely on which company can build the most powerful model. Cost, speed and the ability to deploy AI across millions of tasks are becoming equally important.
GPT-6 Sol and Luna are now available through the OpenAI API, while they are also rolling out across ChatGPT Work and Codex for paid users. Luna will additionally be available to Free and Go users through the desktop app. OpenAI says the models will gradually become available across ChatGPT throughout the day.
With GPT-6 Astra positioned as the company’s high-end model and Sol and Luna targeting cheaper, scalable workloads, OpenAI appears to be building a broader model portfolio designed to cover everything from complex reasoning and coding to routine business automation.
The strategy reflects a simple reality emerging in the AI industry: the next phase of the race may not be about building the biggest model, but about making advanced intelligence cheap enough and efficient enough to be used everywhere.
Leave a Reply