OpenAI on Thursday cut API prices for two of its GPT-5.6 models roughly three weeks after they launched, dropping Luna by 80 percent and Terra by 20 percent in a move that undercuts Anthropic’s mid-tier offering and reframes the frontier-lab price war as a compounding-efficiency story.

Luna’s input pricing falls from $1 to $0.20 per million tokens, with output down from $6 to $1.20. Terra moves from $2.50 to $2 on input and $15 to $12 on output. Sol, the flagship, holds its price but gains a new “Sol Fast” mode offering up to 2.5 times standard processing speed at twice the cost, replacing the older Priority Processing tier. The new Terra output rate sits meaningfully below Claude Sonnet 4.6’s $3 input and $15 output.

The framing OpenAI chose is telling. The company attributed a 20 percent reduction in end-to-end serving cost, and more than 15 percent improvement in token-generation efficiency, to Sol autonomously rewriting production GPU kernels inside a human-led process. That’s the pitch: the model is now paying for its own price cuts. Further reductions are expected as gains compound.

The commercial pressure is legible. Enterprise buyers are scrutinizing AI spend against cheaper Chinese open-weight alternatives, and Anthropic earlier this month priced Claude Opus 5 at half the cost of Claude Fable 5. Both labs have filed confidential IPO prospectuses with the SEC, which turns every pricing decision into a narrative-management exercise for future public-market investors.

Recall the 2014–2016 AWS price-cut cadence, when Amazon used serial reductions to signal scale advantages competitors couldn’t match. OpenAI is running the same play, but with the twist that the efficiency source is the product itself.

Sources