Summary
Paul J Lipsky covers OpenAI’s launch of GPT-6-Sol and GPT-6-Luna, arguing the release could have a bigger everyday impact than GPT-6-Astra due to its pricing. He walks through OpenAI’s claim that GPT-6-Sol costs 50% less than its predecessor, GPT-5.6-Sol, while closing much of the performance gap with the flagship GPT-6-Astra model.
Lipsky reviews benchmark charts covering professional and business workflows, along with the Agents Last Exam benchmark for complex agentic tasks, showing GPT-6-Sol outperforming 5.6-Sol at a lower cost, though still trailing Astra overall. He notes a few exceptions, such as computer-use tasks, where 5.6-Sol technically scores higher but at a steeper price, and highlights that OpenAI carried over Astra’s improved communication style to the new Sol and Luna models.
The video also covers availability, with GPT-6-Sol and Luna rolling out across ChatGPT, Codex, and various subscription tiers, plus a fresh usage-limit reset for subscribers. Lipsky shares his own usage patterns, noting how GPT-6-Astra had been burning through his weekly quota faster than prior models, and expresses hope that GPT-6-Sol can serve as a more sustainable everyday option. He closes by promising further hands-on testing over the following days.
📺 Source: Paul J Lipsky · Published September 22, 2026
🏷️ Format: News Analysis
