Tech leaders at the Sun Valley Conference are increasingly focused on one issue: lowering the cost of AI.
Speaking to CNBC, OpenAI CEO Sam Altman said conversations have shifted from AI adoption to efficiency, with executives asking how to reduce spending while getting better returns on their AI investments.
Altman said this growing demand influenced the development of OpenAI's new GPT-5.6 model family, which includes the flagship Sol model, the balanced Terra model, and the lower-cost Luna model. According to Altman, GPT-5.6 Sol is 54% more token-efficient on agentic coding tasks, making it faster and cheaper for enterprise use.
He added that nearly every business is now evaluating whether its AI spending is delivering enough value.
The comments come as major companies look for ways to cut AI costs. Coinbase CEO Brian Armstrong has experimented with using lower-cost AI models for some engineering tasks, while Vercel CEO Guillermo Rauch has suggested companies combine models from providers like OpenAI, Anthropic, Google Gemini, and others to balance performance with cost.