OpenAI has launched GPT-6 Sol and GPT-6 Luna, two models designed to bring capabilities developed for GPT-6 Astra to lower-cost plans and higher-volume workloads. Sol is positioned for everyday professional work, while Luna emphasizes fast responses and efficient use at scale.

The company says both models improve on their predecessors in professional tasks, factuality, coding, computer use, and alignment. GPT-6 Sol also makes about half as many mistakes as GPT-5.6 Sol in OpenAI's testing, while delivering stronger coding and computer-use performance.

GPT-6 Sol and Luna focus on different workloads

GPT-6 Sol offers more reasoning capability and is intended for users who need a general-purpose model for everyday work. GPT-6 Luna is cheaper and optimized for rapid responses and larger volumes of requests.

OpenAI says the new models outperform their predecessors across its benchmark tests. Sol also matched or exceeded Claude Fable 5.1 on some evaluations. The company compared Sol with Opus 5 on several tests, although Anthropic has since introduced Opus 5.5, which it says outperforms Astra on some coding and knowledge-work benchmarks.

Benchmark comparisons do not provide a complete measure of cost or performance. OpenAI notes that Sol costs half as much per token as Opus 5.5, while Anthropic says its newer model can use fewer tokens per task than Opus 5.

Lower prices and improved prompt caching

OpenAI is reducing token prices for both models compared with their GPT-5.6 counterparts. GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, with both rates reduced by half.

GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens. Its previous input price was $0.20 per million tokens, while output cost $1.20 per million tokens.

Prompt caching is also receiving an upgrade. OpenAI says agents will be able to reuse more context, respond faster, and receive a 90% discount on cached input-token reads. The change could reduce costs for applications that repeatedly work with the same instructions or information.

Astra-style communication and availability

Sol and Luna adopt the communication style OpenAI introduced with Astra. Responses are designed to use less jargon and fewer low-value details, while becoming slightly shorter without losing substance.

OpenAI also reports lower rates of misleading claims about coding work in its alignment testing. That measure is intended to address situations where a model inaccurately describes what it has done or completed.

GPT-6 Sol and GPT-6 Luna are rolling out in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. Luna is also available to Free and Go users in the desktop app. Neither model is currently available in Chat mode.