OpenAI Launches GPT-6 Sol And Luna With 50% Lower API Prices And Fewer Errors

OpenAI has expanded its GPT-6 model family with GPT-6 Sol and GPT-6 Luna, cutting API prices by 50% compared with GPT-5.6 promotional pricing while claiming improvements across factuality, coding, computer use and professional workloads.
 

The two models follow GPT-6 Astra, which OpenAI launched earlier in September as the most capable member of the GPT-6 family. While Astra remains positioned for demanding workloads, Sol and Luna are designed to make GPT-6 capabilities available at lower costs and across higher-volume use cases.

TL;DR

 

• OpenAI launched GPT-6 Sol and Luna with API prices 50% below GPT-5.6 promotional pricing.
• GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, while Luna costs $0.10 and $0.50 respectively.
• OpenAI says Sol made about half as many factual mistakes as GPT-5.6 Sol in an internal evaluation.
• Both models are rolling out across OpenAI products and its API.

OpenAI Cuts GPT-6 Sol And Luna API Prices By 50%

In its official announcement, OpenAI said GPT-6 Sol and Luna were trained using methods similar to GPT-6 Astra, with a focus on bringing improvements in professional work, factuality, coding, computer use and alignment to faster and more affordable models.

Source

The biggest difference is pricing.

GPT-6 Sol costs $2 per one million input tokens and $10 per one million output tokens, compared with promotional GPT-5.6 Sol pricing of $4 and $20 respectively. GPT-6 Luna costs $0.10 per one million input tokens and $0.50 per one million output tokens, down from $0.20 and $1.20. OpenAI describes both models as 50% cheaper overall.

Dataconomy also highlighted the lower pricing as one of the central changes in the launch, noting that OpenAI attributed the reduction to improvements in caching and inference.

The company has also improved prompt caching for GPT-6. OpenAI says cached input-token reads can receive discounts of 90%, while GitHub reported that caching improvements reduced the share of prompt tokens requiring fresh processing by more than 50% across billions of requests to OpenAI models.

GPT-6 Sol Makes Fewer Factual Mistakes, OpenAI Says

Cost is only half the story.

OpenAI says GPT-6 Sol also improved factual reliability. In an internal evaluation based on de-identified conversations where users had previously flagged factual errors, the company found that Sol made "about half as many mistakes as its predecessor."

However, OpenAI cautioned that these conversations were specifically selected because they had produced errors in earlier models and were not representative of typical usage. The company also noted that benchmark evaluations may differ from production ChatGPT because of differences in system prompts, tools and other settings.

TechCrunch similarly reported that OpenAI is positioning improved factuality and fewer coding errors alongside price reductions as major selling points for the new models.

How GPT-6 Sol And Luna Perform

OpenAI is positioning Sol as the more capable option for complex professional, coding and agentic workloads, while Luna targets lower-cost, high-volume tasks.

 

On AutomationBench, OpenAI says GPT-6 Sol at xhigh reasoning effort scored 33.2%, with a cost of $0.27 per task. The company compared that with Claude Opus 5 at maximum effort, which scored 26.9% and cost 11.1 times more per task in the evaluation.

On the DeepSWE software engineering benchmark, GPT-6 Sol at maximum effort scored 68.8%, while GPT-6 Luna scored 66.6%. These figures come from OpenAI's evaluation and should be treated as vendor-reported benchmark results rather than definitive measures of real-world performance.

Where Are GPT-6 Sol And Luna Available?

GPT-6 Sol and Luna are available through OpenAI's API as gpt-6-sol and gpt-6-luna.

OpenAI also said the models are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. GPT-6 Luna is accessible to Free and Go users through the desktop app, while OpenAI said the wider ChatGPT rollout would take place gradually.

Gulf News reported that both models support text and image inputs and have context windows of around 1.05 million tokens. It also framed the launch as part of an increasingly aggressive competition between leading AI developers on cost per useful task rather than raw benchmark performance alone.

That competition was particularly visible on launch day.

TechCrunch and Dataconomy reported that Anthropic released Claude Opus 5.5 roughly 90 minutes before OpenAI unveiled GPT-6 Sol and Luna, putting two major AI model announcements within hours of each other.

For OpenAI, the pitch is clear: Astra remains its flagship for maximum capability, while Sol and Luna are intended to push GPT-6 intelligence into workloads where cost and scale matter just as much as raw performance.