OpenAI lowers prices of two AI models as cost concerns grow

OpenAI announced Thursday that it will reduce the prices of two of its latest artificial intelligence models, the GPT-5.6 Terra and GPT-5.6 Luna, approximately three weeks after their launch.

The company faces pressure to serve a more cost-conscious customer base, where companies are less inclined to introduce expensive models without a clear understanding of the return on investment. The company is also working to fend off competition from Chinese startups and tech giants. google and microsoftpromotes a cost-effective model.

OpenAI launched three models as part of the GPT-5.6 series, including Sol, the most powerful product, Terra, the mid-tier model, and Luna, the fastest product.

The company announced on Thursday that it was cutting Terra’s price by 20% to $2 per million input tokens and $12 per million output tokens. Luna’s cost will be reduced by 80% to 20 cents per million input tokens and $1.20 per million output tokens. The price of Sol will not change.

“Our strategy remains focused on improving both the capabilities and efficiency so that each generation of intelligence can do more work at lower cost,” OpenAI said in a release.

OpenAI sparked an AI boom with the launch of ChatGPT in 2022, prompting companies across the U.S. to rush to adopt the technology and encourage adoption within their workforces. The era of so-called token maxing was born, where employers encouraged employees to use as much AI as possible without worrying about cost.

But as AI-related costs continue to balloon, sometimes reaching billions of dollars, many companies are actively trying to rein in their spending. This is a change that AI developers are watching closely, especially as China’s open weight model has quickly caught up with leading proprietary products.

Openweight AI models serve as a cost-effective option for businesses because users can download, modify, and run them on their own infrastructure. Earlier this month, Chinese startup Moonshot AI released an open-weight model called Kimi K3 that outperformed state-of-the-art U.S. products on some industry benchmarks, sparking a swift reaction from Silicon Valley.

Following the announcement of Moonshot, OpenAI’s biggest rival, humanhas released a new model called Claude Opus 5, touting it as the best performing and most cost effective product for many use cases. It’s half the price of the Claude Fable 5, Anthropic’s higher-end model introduced in June, but it performs just as well across coding and knowledge work tasks.

Microsoft CEO Satya Nadella reiterated the company’s cost-effective model during a quarterly earnings call with investors on Wednesday, after announcing what he called an inexpensive but high-performance cybersecurity model earlier this week.

Google also announced three new models this month aimed at undercutting competitors in costs. The company says the Gemini 3.6 Flash, the most powerful of its new products, has a lower price per task than the Kimi K3 and other Chinese models.

clock: OpenAI CEO Sam Altman meets with White House Chief of Staff Susie Wiles this week