Power and Savings: OpenAI Drops GPT-6 Sol and Luna to Slash Coding Costs and Hallucinations

Following the release of its flagship GPT-6 Astra system earlier this month, OpenAI expanded its sixth-generation model lineup by introducing updated versions of its lightweight Sol and Luna models. Company updates state that these smaller models bring high-tier intelligence to broader developer tools while cutting operational expenses and computing overhead.

OpenAI positions each model size for distinct workload demands. Sol handles complex technical projects like software development, code reviews, and data analysis. Meanwhile, Luna targets basic administrative chores, managing high-volume tasks like summarizing long documents, extracting unstructured text, and answering user queries.

Alongside functional updates, OpenAI slashed API access pricing across its entire 6-series lineup. The updated 6-series models cost roughly half as much as previous builds, a price drop driven by backend caching improvements and optimized system inference.

OpenAI also claims these refreshed models cut error rates significantly when writing code or answering factual questions. According to internal accuracy benchmarks using anonymized real-world conversations, GPT-6 Sol makes half as many factual mistakes as its predecessor. These efficiency gains allow developers to get reliable, high-tier output at a fraction of past API costs.

The launch fuels ongoing competition with rival research lab Anthropic. OpenAI claims GPT-6 Sol and Luna beat Anthropic’s top platforms, including Fable and Opus, on standard technical benchmarks. Notably, Anthropic released an updated version of Opus 5.5 just 90 minutes before OpenAI’s announcement, highlighting the fast pace of rival model rollouts.

The new Sol and Luna builds are available immediately across ChatGPT Work and Codex for paid account tiers, as well as through the standard ChatGPT API. Free and Go users gain access to Luna inside the desktop and mobile apps as OpenAI rolls out the update globally.

Slashing operational costs while improving factual accuracy changes the math for enterprise development teams. High API costs previously forced software teams to limit how often their background agents ran checks, analyzed logs, or reviewed code. Cutting token prices by half gives engineers room to run heavy background tasks continuously without blowing through monthly infrastructure budgets.

As model providers compete fiercely on price and speed, building reliable software gets cheaper and far more accessible for everyday developers. Teams no longer have to choose between cheap models that hallucinate or expensive models that burn cash quickly. OpenAI’s mid-tier engines prove that smaller, focused software models can handle heavy lifting while keeping cloud overhead under control.

Furthermore, these price drops create a snowball effect across the entire developer ecosystem. When foundational platform costs drop by 50%, startups can experiment much faster, testing aggressive feature ideas that were previously too expensive to run at scale. Independent developers can build, test, and ship full-featured software products with minimal upfront capital. At the same time, enterprise development teams can expand their background processing pipelines, running automated code analysis, real-time security scanning, and continuous integration checks on every single commit. By making intelligent models faster, cheaper, and more accurate, OpenAI is forcing the entire industry to rethink how software gets built and deployed in modern tech stacks.