# OpenAI Cuts GPT 5.6 Luna Price By 80%, GPT 5.6 Terra Price By 20% > Source: > Published: 2026-07-30 17:57:10+00:00 OpenAI appears to be responding to the threat of cheaper Chinese open-source models with some price cuts of its own. Sam Altman announced on Thursday that OpenAI is slashing the price of GPT-5.6 Luna, the fastest and cheapest model in its [GPT-5.6 lineup](https://officechai.com/ai/openai-launches-gpt-5-6-sol-beats-mythos-on-terminalbench/), by 80%. Luna now costs 20 cents per million input tokens and $1.20 per million output tokens, down from its launch pricing of $1 and $6. Terra, the mid-tier model in the family, gets a smaller but still meaningful cut, dropping 20% to $2 per million input tokens and $12 per million output tokens. Sol, the flagship, keeps its pricing unchanged at $5 input and $30 output. Altman framed the move as part of a broader push to give developers a better price to intelligence ratio at every tier, rather than just shipping one flagship model and letting everything else trail behind it on cost. Along with the cuts, Sol is getting a Fast mode in the API that runs up to 2.5 times quicker at double the price, replacing the Priority Processing option that existed before it. The intelligence stays the same, OpenAI says, only the speed and the bill change. The timing is hard to separate from what has been happening to OpenAI’s market share over the past several months. Chinese labs have been undercutting Western frontier models on price by wide margins, and enterprises have noticed. [DeepSeek’s V4 family](https://officechai.com/ai/deepseek-v4-pro-deepseek-v4-flash-benchmarks-pricing/), for instance, launched with output pricing a fraction of what OpenAI and Anthropic charge for comparable work, and OpenRouter data has shown the combined token share of Google, OpenAI, and Anthropic models falling from around 70% a year ago to closer to 30% today. A good chunk of that traffic has moved to open-weight Chinese models that show up on lists of the [cheapest AI models](https://officechai.com/ai/cheapest-ai-models/) in the world, many of them priced well under a dollar per million tokens. Luna’s new rate puts it in that same conversation. At 20 cents input and $1.20 output, it undercuts most of what Anthropic and Google currently charge for their own budget-tier models, and it lands close to where several Chinese open-weight options already sit. Terra’s new price of $2 input and $12 output is a smaller move by comparison, but it still puts pressure on Anthropic’s mid-tier Claude Sonnet, which is priced higher on both input and output. Whether this is enough to pull enterprise workloads back from cheaper open-weight alternatives is a separate question from whether it makes Luna and Terra better deals on their own terms. The chart Altman shared alongside the announcement shows GPT-5.6 Luna sitting well ahead of the pack on intelligence per dollar, with models like Claude Sonnet, Claude Opus, and Gemini clustered at higher price points for similar or lower scores on the Artificial Analysis Intelligence Index. That kind of chart tends to look better right after a price cut than it does a few months later, once competitors respond. Anthropic and Google have both cut prices on their own lower-tier models over the past year whenever OpenAI has made a similar move, and there is little reason to expect this round to be different. For developers already building on GPT-5.6, the practical effect is straightforward: workloads currently running on Terra get cheaper without any code changes, and anything routed to Luna for high-volume, latency-sensitive tasks gets meaningfully cheaper still. The bigger question is whether OpenAI keeps making this kind of adjustment every few weeks as new competition lands, or whether Thursday’s cut is a one-time correction to a launch price that priced Luna out of the conversation it was supposed to win.