OpenAI introduced GPT-6 Sol and Luna for programming with a cheap API

OpenAI has introduced GPT-6 Sol and GPT-6 Luna — new models in the GPT-6 family, designed for professional and everyday tasks. At the same time, the company reduced the cost of using them via API by 50% compared to the promotional rates of the GPT-5.6 Sol and Luna models.

GPT-6 Sol is intended for more complex tasks, including programming and code-related work. GPT-6 Luna is designed for tasks with clear objectives, such as summarizing documents, extracting information, and answering questions.

The cost of GPT-6 Sol is $2 per 1 million input tokens and $10 per 1 million output tokens. For GPT-6 Luna, the price is $0.10 per 1 million input tokens and $0.50 per 1 million output tokens.

OpenAI attributes the cost reduction to improvements in caching and inference. The company has also updated its prompt caching mechanism: developers can receive a discount of up to 90% on processing previously cached input tokens.

According to OpenAI, GPT-6 Sol makes factual errors approximately twice as rarely as GPT-5.6 Sol in the company's internal testing. At the same time, OpenAI clarifies that the evaluation is based on anonymized real dialogues where users previously pointed out model errors, and does not reflect typical usage levels.

The new model has also received improvements in programming. On the DeepSWE v1.1 test, GPT-6 Sol scored 68.8% at maximum effort level, while GPT-6 Luna scored 66.6%. OpenAI notes that Sol requires approximately 80% less cost per task compared to the comparable Claude Fable 5 model.

GPT-6 Sol and Luna have also improved their performance with computer and professional tasks. However, GPT-6 Astra remains OpenAI's most powerful model for tasks requiring maximum performance.

GPT-6 Sol and GPT-6 Luna are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. Through the API, the models are available under the names gpt-6-sol and gpt-6-luna. GPT-6 Luna is also available to Free and Go users in the desktop application. In ChatGPT, the models are rolled out gradually.

Source: openai.com