OpenAI’s ChatGPT 5.6 Sol Cuts AI Serving Costs by 20 Percent

OpenAI unveiled ChatGPT 5.6 Sol, a model focused on efficiency rather than pushing raw capability.

GPU kernel optimizations and speculative decoding cut serving costs by 20% and improved token generation efficiency by 15%.

The model was further refined using GPT-5.6 itself, reflecting a self-optimization approach to model development.

The move highlights a broader industry shift toward lower costs, better energy use, and more sustainable AI deployment amid safety and policy tensions.