Advancing the price-performance frontier with GPT‑5.6

Simon Willison's Weblog is reporting major price reductions from OpenAI for its G-P-T five point six model line. The Terra variant dropped twenty percent, while Luna saw an eighty percent price reduction. Luna is now twenty cents per million tokens for input and one point two dollars per million for output. OpenAI achieved these savings by using G-P-T five point six Sol to optimize inference efficiency and rewrite production kernel code in Triton and Gluon, open-source G-P-U languages. The optimization effort reduced end-to-end serving costs by twenty percent. Luna's new pricing reshapes the lower-cost A-I market—it now costs about one-fifth as much as Anthropic's Claude Haiku four point five for input tokens.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton