R to @OpenAI: Making advanced intelligence more abundant and affordable is central to our mission…

TL;DR - OpenAI announced API price cuts for its Luna and Terra models and faster performance for Sol, funded by efficiency gains that GPT-5.6 Sol itself helped produce. It matters because it frames self-improving model-assisted systems work as a direct lever on inference cost and serving throughput.

• OpenAI says it applied GPT-5.6 Sol to optimize its own serving stack, i.e. the model contributed to making itself cheaper to run.
• Claimed ~20% lower serving costs from production GPU kernel improvements.
• Claimed 15%+ better token-generation efficiency from improved speculative decoding.
• Gains are passed through as lower API prices (Luna, Terra) and faster latency (Sol); no benchmark or quality-impact data was provided.