ChatGPT-6 Sol and Luna are here: Pricing and benchmarks data

OpenAI has introduced two new ChatGPT models, GPT-6 Sol and Luna, in a move to challenge Anthropic's Claude Opus 5.5. The new models were announced on Tuesday, just hours after the launch of Claude Opus 5.5. According to OpenAI's blog post, these models offer improved accuracy and cost-efficiency compared to their predecessors.

GPT-6 Sol and Luna are essentially scaled-down versions of GPT-6 Astra, which was released earlier in September. They are designed to cater to different work scales than the ones handled by Astra. OpenAI trained the new models using similar methods as those used for GPT-6 Astra, aiming to bring the state-of-the-art performance in professional work, factuality, coding, computer use, and alignment to faster and more affordable models.

These new models are now available for Plus, Pro, Enterprise, Business, and Edu customers in Work and Codex, while Free users can access Luna in the ChatGPT desktop app. However, neither model is currently available in basic Chat. The pricing for the new models is set at $2 for input and $10 for output for GPT-6 Sol, and $0.10 for input and $0.50 for output for GPT-6 Luna per 1 million tokens. This represents a 50 percent decrease in cost compared to GPT-5.6 Sol and Luna.

OpenAI claims that GPT-6 Sol and Luna have dramatically reduced factual errors compared to the older 5.6 variants, making half as many errors as before. The company asserts that this improvement was achieved within a single financial quarter. To substantiate this claim, OpenAI provided benchmark data showing that GPT-6 Luna outperformed the 5.6 release by 5.4 percent, while costing 58 percent less per task. The new GPT-6 models outperformed their predecessors in all other benchmarks, as stated in OpenAI's blog post.

It is worth noting that Ziff Davis, Mashable's parent company, filed a lawsuit against OpenAI in April 2025, alleging infringement of Ziff Davis copyrights in training and operating its AI systems.