OpenAI's small GPT-5.6 model fell 80% in price in July. By September, the flagships followed

The cuts, in order

The price timeline starts on June 26 with GPT-5.6 Sol at $5 input and $30 output per million tokens. On July 30 OpenAI cut its small GPT-5.6 Luna by 80%, to $0.20 and $1.20. Anthropic made a temporary Sonnet 5 launch price of $2 / $10 permanent on August 10. On September 22 OpenAI shipped GPT-6 Sol at $2 / $10 and Luna at $0.10 / $0.50, while Anthropic released Opus 5.5 at $4 / $20. A week later GPT-6.1 Sol halved its cache-read price again.

Five guesses, one puzzle

He offers five reasons. Substitutes make models interchangeable. Providers hosting open models set a floor near marginal cost. Anthropic says Opus 5.5 needs less compute. The two labs are answering each other. And cache reads make up 96 to 99 percent of agentic coding spend, so cutting that line moves the whole bill.

What he cannot explain is timing. Both companies are expected to list, and cutting revenue looks odd in an S-1. He also notes that OpenAI cut the Pro-tier Codex allowance by 50%, so some of the cheapness was taken back elsewhere.

The thread: cheaper because the alternatives are sitting right there

"Competition drives prices down. We tried Kimi and GLM models after hitting 'cybersecurity' blocks on OpenAI and Anthropic. We found them to be much cheaper to access and helped us complete our work without 'cybersecurity' refusals blocking us." — a commenter whose team switched models after refusal blocks

Competition is also what worried others about the sums.

"The frontier labs were already burning money on more expensive price models, and competition wouldn't have reduced their capex proportionately." — a commenter who reads the price cuts as a warning

That anxiety, not the price table, is the part the post leaves open. If a price cut is competition forced by free open models, the margin the S-1 would need is the one being given away.