OpenAI Just Made Frontier AI 80% Cheaper to Run

Three weeks after launching its GPT-5.6 family of models on July 9, 2026, OpenAI cut the price of using them sharply. On July 30, the price of GPT-5.6 Luna, the smallest and cheapest model, fell by 80%, from $1 to $0.20 per million input tokens and from $6 to $1.20 per million output tokens. Terra, the mid-tier model, dropped 20%, from $2.50 to $2 for input and from $15 to $12 for output. The cuts went live in OpenAI's API and on Amazon's AWS Bedrock the same day.

The top-tier Sol model initially stayed at $5 and $30, though OpenAI added a "Fast" mode that runs up to 2.5 times faster at twice the price. Then, on August 21, it cut Sol too. "As we continue to push the frontier of capabilities while improving efficiency, we're dropping API and credit pricing of GPT-5.6 Sol by over 20% for the next 3 months," the company said. Sol now costs $4 per million input tokens and $20 per million output tokens, at least through November 21, 2026, with the discount also covering Fast mode, long-context requests, and Batch and Flex processing.

What a "token" price actually means

For anyone outside the industry, token prices don't mean much on their own. A token is a chunk of text, roughly three-quarters of an English word on average, so a million tokens is somewhere around 750,000 words. Developers pay for both the text they send to the model (input) and the text it writes back (output).

At Luna's new price, processing a million tokens of input costs 20 US cents, roughly ₹19. With Batch and Flex processing, which handle work that doesn't need an instant answer at half the standard rate, it falls to $0.10 for input and $0.60 for output. That is cheap enough that tasks like summarising thousands of support tickets, sorting documents, or powering a customer-service chatbot become affordable for very small teams.

Why OpenAI is cutting prices

CNBC reported that the cuts come as companies grow more sensitive to the cost of running AI at scale. Early experiments with AI are cheap; running it across an entire business, millions of times a day, is not, and price has become a key factor when companies choose which model to build on. OpenAI also faces strong competition from Google, Anthropic, and a growing number of open-source models that companies can run themselves, such as those from River AI, which raised $1.1 billion to build around open models.

Price cuts like these usually reflect a mix of competitive pressure and genuine efficiency gains, as AI companies find cheaper ways to serve the same model. Consumers shouldn't expect any change: OpenAI said ChatGPT Plus, Pro, and Business subscription prices are unaffected.

Why it matters for India

India has one of the world's largest developer communities and a fast-growing startup ecosystem, and for many Indian teams, dollar-denominated AI pricing has been a real barrier. Each cut like this widens the range of products that make financial sense to build, from vernacular-language assistants and tutoring apps to tools for small businesses.

Cheaper AI doesn't guarantee it gets used well. But it does lower the barrier for who gets to build with it. A year ago, this level of capability was priced for large enterprises. Now it is priced for almost anyone with an idea and a laptop.