OpenAI GPT-5.6 price cuts target companies watching AI costs
OpenAI cut prices for GPT-5.6 Terra and Luna weeks after launch as enterprises scrutinize AI spending and cheaper rivals gain ground.
By Jordan Bell · Startups & Deals Reporter
· 3 min read
OpenAI GPT-5.6 price cuts are arriving less than a month after the models went public, a sign that AI providers are fighting harder on cost as businesses question how much they should spend on generative AI. OpenAI said Thursday it is lowering prices for GPT-5.6 Terra and GPT-5.6 Luna, while leaving its top-end GPT-5.6 Sol model unchanged.
The move matters for investors watching the AI trade because model pricing affects how quickly companies adopt AI tools, how much revenue developers can capture, and how aggressively rivals compete. CNBC reported that OpenAI is trying to serve customers that have become more careful about AI budgets while also responding to pressure from Chinese startups and large tech companies including Google and Microsoft.
What are OpenAI's GPT-5.6 price cuts?
OpenAI said it is cutting GPT-5.6 Terra by 20%, bringing the model to $2 per million input tokens and $12 per million output tokens. Tokens are the small units of text that AI systems process, so input tokens are what a user sends to the model and output tokens are what the model generates back.
OpenAI said GPT-5.6 Luna, the fastest model in the series, will become 80% cheaper. Luna will cost 20 cents per million input tokens and $1.20 per million output tokens.
The GPT-5.6 lineup includes three models: Sol, described by CNBC as the most powerful option; Terra, the mid-tier version; and Luna, the speed-focused model. OpenAI released the series publicly roughly three weeks before the price cuts.
“Our strategy remains focused on advancing both capability and efficiency so each generation of intelligence can accomplish more work at a lower cost,” OpenAI said in a release.
Why AI model prices are becoming a bigger issue
OpenAI helped set off the current AI boom when it launched ChatGPT in 2022. CNBC reported that U.S. companies then rushed to put the technology into workers’ hands, often encouraging heavy use before costs became a central concern.
That phase has changed. CNBC reported that AI bills have grown sharply, in some cases reaching billions of dollars, pushing companies to limit spending and demand clearer returns before rolling out expensive models more broadly.
Cheaper open-weight models have added to that pressure. Open-weight models can be downloaded, modified and run on a company’s own systems, which can make them a lower-cost alternative for some users.
CNBC reported that Chinese startup Moonshot AI released an open-weight model called Kimi K3 earlier this month and that it outperformed leading U.S. models on some industry benchmarks. The report said the release drew a quick response from Silicon Valley.
OpenAI is not the only company emphasizing efficiency. CNBC reported that Anthropic followed Moonshot’s announcement with Claude Opus 5, which it promoted as both its strongest and most cost-effective model for many uses. Anthropic said Claude Opus 5 costs half as much as Claude Fable 5, an advanced model released in June, while performing comparably on coding and knowledge work tasks.
Microsoft CEO Satya Nadella also repeatedly pointed to cost-effective models during the company’s quarterly earnings call Wednesday, according to CNBC, after Microsoft introduced what it described as a low-cost cybersecurity model earlier in the week. Google, meanwhile, debuted three models this month aimed at lowering customer costs, with the company saying Gemini 3.6 Flash is cheaper per task than Kimi K3 and other Chinese models.
This story draws on original reporting from CNBC.
