OpenAI updates its AI lineup with GPT-6 Sol and GPT-6 Luna models

Add Prompolitinform to your Preferred Sources

You will see our news more often on Google and find important Prompolitinform stories faster.

COMMUNICATIONS AND IT 23.09.2026 / author:
OpenAI updates its AI lineup with GPT-6 Sol and GPT-6 Luna models

API costs for the new models have been halved compared to GPT-5.6 pricing

OpenAI has introduced new artificial intelligence models, GPT-6 Sol and GPT-6 Luna, aimed at increasing efficiency and reducing costs. The developers have reduced API prices for these models by 50% compared to the previous GPT-5.6 version. This was reported by OpenAI, according to the portal PromPolitInform.

The new versions complement the company’s product line, which is based on the GPT-6 Astra model technology. The update focuses on improving professional performance, particularly in programming, data analysis, and computer interface usage. Technical optimizations, especially in caching mechanisms, have significantly increased AI response speed and reduced infrastructure load during the processing of long queries.

Economic benefits and accessibility

Thanks to technological changes, the cost of using the API for Sol and Luna models has become much more attractive to the business sector. The price per million input tokens for GPT-6 Sol has decreased from 4 to 2 dollars, and for output from 20 to 10 dollars. The Luna model is even more affordable: costs for input data decreased from 0.20 to 0.10 dollars per million tokens, and for output from 1.20 to 0.50 dollars.

Productivity in work tasks

Testing on the AutomationBench platform showed that the GPT-6 Sol model is capable of executing complex business processes much more efficiently than competitive counterparts at lower financial costs. Additionally, the developers have refined the systems’ communication style, making AI responses more concise and free of unnecessary technical jargon.

Caching optimization and development

The update includes cache diagnostic tools that allow developers to better control resource consumption while working with agents. For example, the implementation of explicit breakpoints allows for the selection of caching prefixes, which facilitates context reuse and minimizes the need for data re-processing.

User access

Starting today, the models are available in ChatGPT Work and Codex services for Plus, Pro, Business, Enterprise, and Edu plan users. Free tier and Go users can access GPT-6 Luna via the desktop application, with a phased rollout of the models continuing throughout the day to ensure system stability.

Illustrative photo: Dietmar Rabich / CC BY-SA 4.0

Read also: Donald Trump to appoint an AI czar