OpenAI launches GPT-6 Sol and Luna: 50% cheaper and more powerful than competitors
OpenAI has introduced two new models—GPT-6 Sol and GPT-6 Luna—which are already available via API and are beginning to be integrated into the Codex and ChatGPT Work products. The main innovation is a 50% reduction in API cost compared to the previous GPT-5.6 version.
These models are designed for different tasks: Sol is suitable for content creation, while Luna is for scaling and complex business processes. Their high efficiency is achieved through improved prompt caching, which allows agents to perform less unnecessary background processing, making responses faster and cheaper.
Tests show that GPT-6 Sol and GPT-6 Luna demonstrate impressive results. For example, in coding tasks (DeepSWE), Sol is almost comparable to Claude Fable 5 at a cost that is 80% lower. This confirms that users get more intelligence for significantly less money.
Why it matters
- —A significant 50% reduction in API costs makes advanced AI capabilities accessible to a wider range of users.
- —The models demonstrate high performance in critical tasks such as coding and business process automation.
- —Improved caching and process optimization boost the overall efficiency of AI agents.
Key facts
- GPT-6 Sol and GPT-6 Luna are available with a 50% reduction in API price compared to GPT-5.6.
- In the DeepSWE v1.1 test, GPT-6 Sol approached Claude Fable 5 at a cost that is 80% lower.
- In AutomationBench, GPT-6 Sol outperformed competitors, showing a claimed cost 88% lower.
- Improved prompt caching reduces redundant data processing, increasing speed and lowering cost.
The full text is in the original source. Here we provide a brief summary and key facts.
