OpenAI has announced the new GPT-6 Sol and Luna models following the GPT-6 Astra. These new models bring the improvements offered by Astra to users at more accessible prices in areas such as professional workflows, coding, and computer use. According to OpenAI, these models were prepared using development formulas similar to those used with Astra. They aim for higher accuracy and smoother alignment performance, especially in professional processes.

Innovations with GPT-6 Sol and Luna

The GPT-6 Sol model offers a structure optimized for daily tasks and comes at a more affordable cost compared to Astra. This model particularly excels in reasoning capabilities.

GPT-6 Luna stands out as a more cost-effective option. Luna is specifically optimized for tasks requiring fast response times and high throughput.

The new models outperform their predecessors in OpenAI’s benchmark tests. GPT-6 Sol makes approximately half as many errors as the GPT-5.6 Sol model.

Offering significant improvements in coding and computing, Sol achieves or surpasses Claude Fable 5.1 scores in several tests. While OpenAI compares the Sol model to Opus 5, Anthropic’s new Opus 5.5 model outperforms Astra in several coding benchmarks.

Both Sol and Luna models adopt the new communication style introduced with Astra. This style offers a simpler, more concise language with less jargon and fewer unnecessary details.

Responses are kept shorter without losing content. Additionally, new models significantly reduce the rate of incorrect statements in coding tasks.

Performance and Usability

OpenAI has reduced token prices for Sol and Luna models compared to previous 5.6 versions. Improvements have also been made to prompt caching processes.

Response times are faster thanks to agents using context more efficiently. A 90% reduction is achieved in cached input token reads.

GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens. These prices represent a 50% reduction compared to previous versions.

GPT-6 Luna is offered at a cost of $0.10 per million input tokens and $0.50 per million output tokens. These numbers are significantly lower than the previous prices of $0.20 and $1.20.

GPT-6 Sol offers a cost per token that is half as cheap as the Opus 5.5 model. However, Anthropic states that a direct comparison is difficult, noting that Opus 5.5 uses fewer tokens per task.

GPT-6 Sol and Luna are currently available to Plus, Pro, Business, Enterprise, and Edu users via ChatGPT Work and Codex. GPT-6 Luna is also available for Free and Go users in the desktop application.

The new models are not currently available in Chat mode. How do you think the price and performance stability offered by these new models will change your AI usage habits?

OpenAI announced the new GPT-6 Sol and Luna models. Focused on everyday tasks and high speed, these models come with affordable options.