OpenAI has launched two new AI models, GPT-6 Soul and GPT-6 Luna, expanding its GPT-6 model family with a focus on reducing costs, improving accuracy, and lowering error rates compared with previous versions.
اضافة اعلان
The launch follows the introduction of GPT-6 Astra earlier in September, which OpenAI described as its most powerful and capable model to date, designed to handle complex tasks including coding and computer use.
OpenAI said Soul and Luna were trained using methods similar to those used for Astra, aiming to bring improvements in professional work, factual accuracy, coding, computer use, and alignment to faster and more affordable models.
“GPT-6 Astra introduced a new generation of intelligence, and these models extend its benefits by making that intelligence more efficient and accessible,” the company said.
Earlier versions of Soul and Luna were released earlier this year, with the two models designed for different levels of performance within OpenAI’s ecosystem.
Soul is primarily designed for more complex tasks, including coding and work requiring advanced reasoning. Luna, meanwhile, is aimed at high-volume office tasks with clear objectives, such as summarizing documents, extracting
information, and answering quick questions.
Lower Costs and Higher Accuracy
OpenAI said one of the key improvements in the new generation is a roughly 50% reduction in API costs compared with GPT-5.6 Soul and Luna.
Based on the announced prices per million tokens, GPT-6 Soul costs $2 per million input tokens and $10 per million output tokens.
GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens.
OpenAI attributed the lower prices to improvements in caching and inference technologies, allowing the models to operate more efficiently at a lower cost.
Fewer Errors and Greater Accuracy
The company also said the two new models are better at providing factually accurate answers and have lower coding error rates than previous versions.
In an internal accuracy evaluation based on anonymized real-world conversations in which users had reported model errors, OpenAI said GPT-6 Soul made about half as many errors as its predecessor.
The company added that Soul now offers a level of reliability close to Astra, but at a significantly lower cost.
According to data published by OpenAI, GPT-6 Luna outperformed version 5.6 by 5.4% on an automation-task performance test, while the cost of completing each task fell by 58%.
OpenAI also presented results from other performance tests that, according to the company’s data, showed the new GPT-6 models outperforming previous versions across the evaluations it conducted.
The launch comes only a few months after the release of GPT-5.6 Soul and Luna. According to OpenAI’s figures, the company has significantly reduced error rates while also lowering operating costs in a relatively short period.
Growing Competition With Anthropic
The launch comes amid intensifying competition between OpenAI and Anthropic.
Anthropic announced a new version of its Opus 5.5 model about 90 minutes before OpenAI announced Soul and Luna.
OpenAI said its two new models outperform Anthropic’s advanced models in several tasks, including its Fable and Opus models, based on performance tests published by the company.
Where Are the New Models Available?
The new versions of Soul and Luna are available in ChatGPT Work and Codex for users on Plus, Pro, Enterprise, Business, and Edu plans. They are also available through OpenAI’s API.
Users on the Free and Go plans can access Luna through the ChatGPT desktop app.
However, the two models were not yet available in ChatGPT’s main chat interface at the time of the announcement. OpenAI said it would gradually roll them out through the app and website throughout the day.
