OpenAI launches GPT-6 Sol and Luna: cheaper models with fewer errors

OpenAI continues to expand its GPT-6 product lineup, announcing the launch of two new models: Sol and Luna. The company claims both variants bring significant improvements in operating costs and the accuracy of generated responses, according to TechCrunch.

What's new with Sol and Luna

According to available information, the two models are "cut from the same cloth as Astra," as TechCrunch put it, suggesting that OpenAI relied on an architecture or set of training principles similar to those behind the previously released Astra model. This approach appears to point to a strategy of technological consolidation, with the company refining already-tested solutions rather than reinventing its architecture from scratch each time.

From a business user's perspective, the most significant changes seem to be tied to cost efficiency. A lower cost per query or per processed token could make these models far more attractive to companies running AI applications at scale, where API expenses can quickly add up to substantial sums.

Impact on the enterprise AI market

A reduced error rate is another key point highlighted by OpenAI. For sectors such as financial services, healthcare, or legal support—where the accuracy of AI-generated information is critical—improved precision could open up new deployment opportunities in areas where companies have so far been hesitant due to the risks associated with language model errors.

The launch comes at a time when competition in the large language model market has intensified considerably, with players like Google, Anthropic, and Meta rolling out constant updates. Differentiating through cost and reliability, rather than raw capability alone, appears to be an increasingly clear strategic direction for OpenAI.

It remains to be seen how Sol and Luna will perform in independent tests and how quickly they will be adopted by companies that already rely on OpenAI's infrastructure for critical operations. For now, public information remains limited, with further technical details expected in the coming period.