OpenAI has released two new artificial intelligence models, GPT-6 Sol and GPT-6 Luna, designed to offer high-level reasoning and programming capabilities at half the cost of previous versions. While these cheaper tools will likely accelerate the integration of AI into schoolwork, emerging educational research warns that unsupervised use of AI leads to lower retention and "cognitive surrender" among students.
What Happened
According to the OpenAI announcement, the tech company is expanding its GPT-6 lineup with "Sol" and "Luna," two models that bring advanced intelligence to users at a much lower cost. The company cut API prices for these models by 50% compared to previous promotional pricing, resulting in a large reduction in the cost of automated tasks. On benchmark tests like Zapier's AutomationBench, GPT-6 Sol outperformed older, more expensive models like Claude Opus 5 at a fraction of the cost. OpenAI also claims these new models are far more reliable, with GPT-6 Sol making roughly half as many factual mistakes as its predecessor.
The Bigger Picture
While tech companies lower prices and promise higher accuracy, educators face a more complex reality. A study on student help-seeking behaviors reveals that when students use AI chatbots, they often fall into "GenAI dependence" rather than using the tool for learning support. This dependency has real consequences for academic achievement. A study tracking math students found that the availability of ChatGPT led to a 26.9% decline in study time for college students and a 31.3% drop for high schoolers. More concerningly, this shortcut reduced their chances of passing subsequent proctored exams by 25%, indicating a "population-level indicator of cognitive surrender."
Even as AI models become cheaper, school systems struggle to adopt them safely. School districts already spend billions of dollars on AI tools but struggle to vet which ones actually improve learning. Many school districts operate on thin margins. For example, the Duluth Public Schools tech department runs on a limited budget that forces them to delay replacing student Chromebooks. When schools do try to buy into expensive AI programs, such as a $3.2 million state-funded safety pilot in North Carolina, communities often reject them due to privacy, security, and data tracking concerns.
What This Means for Families
For parents and teachers, cheaper AI means children will have easier, more constant access to these tools, whether through homework helpers or coding assistants. However, using AI to generate quick answers does not equal learning. A study on AI roles in programming courses found that while using AI as a "problem solver" helped students complete immediate homework tasks, it did not improve their performance on exams. Even when AI is used as a tutor, it does not automatically guarantee better grades. Computer science teachers frequently express concern over student overreliance on AI, which can undermine independent critical thinking. To make these tools effective, students must learn how to guide the AI. Researchers found that integrating Socratic questioning techniques with AI tutoring helped students develop higher-order computational thinking and metacognitive awareness, though it still did not raise raw exam scores.
What You Can Do
- Encourage children to use AI as a brainstorming partner to explain confusing concepts, rather than using it to find the final answer to a math or coding problem.
- Teach students to ask AI questions like "Can you guide me through this step-by-step without giving me the answer?" to prevent cognitive surrender.
- Ask your school board how they vet AI tools for data security, ensuring they protect student data before integrating new classroom software.