OpenAI has cut the price of its GPT-5.6 Luna by 80%, becoming the latest major AI company to respond to growing pressure over the cost of deploying advanced models. Within days, DeepSeek added to that pressure by releasing V4-Flash, a model researchers have identified as the cheapest major AI model currently available.
Instead of focusing only on building smarter models, these AI companies are increasingly trying to make their models cheaper to run, especially as businesses are increasingly paying closer attention to AI costs.
OpenAI Makes One of Its Biggest Price Cuts Yet
OpenAI announced the new pricing on July 30, just three weeks after launching the GPT-5.6 family. But the ChatGPT-maker then reduced the price of its entry-level GPT-5.6 Luna model by 80% and cut the cost of its mid tier GPT-5.6 Terra model by 20%. Pricing for GPT-5.6 Sol, its flagship model, remains unchanged.
These changes mean Luna now costs $0.20 per million input tokens instead of $1, while output tokens have dropped from $6 to $1.20 per million. For Terra, its input price fell from $2.50 to $2 per million tokens, while output pricing dropped from $15 to $12. OpenAI said the reductions were made possible by efficiency improvements that allow it to deliver more intelligence at a lower cost.
DeepSeek Raised the Stakes
Only days later, DeepSeek introduced V4 Flash, a model that research firm Artificial Analysis identified as the lowest cost major AI model currently available.
According to the firm’s analysis, V4 Flash costs $0.14 per million input tokens and $0.28 per million output tokens. It also averages about three cents per benchmark test, making it significantly cheaper to run than OpenAI’s GPT-5.6 Sol, Anthropic’s Claude Fable 5 and several other leading models while still delivering competitive performance.
DeepSeek’s latest release adds to a growing trend among Chinese AI companies that are pushing down inference costs while steadily improving model performance. Companies including Alibaba, Moonshot AI and Z.ai have all introduced increasingly capable models that cost far less than many Western alternatives this year.
AI Companies Are Competing on Cost
For much of the past two years, AI companies have tried to outperform one another by releasing larger and more capable models. That competition has not disappeared, but pricing is becoming just as important.
Businesses are now paying much closer attention to how much AI costs to deploy at scale, especially as models are used in customer service, coding, research and enterprise software. Lower inference costs can make the difference between an AI feature being commercially viable or too expensive to operate.
OpenAI’s latest price cuts reflect this reality. While the company said the reductions were driven by efficiency gains, they also arrive as lower-priced rivals continue to narrow the gap with leading American AI labs.
The Next Battle May Be Profitability
The latest announcements from OpenAI and DeepSeek show that the AI race is entering a new phase. Performance still matters, but companies must now prove they can deliver powerful models without driving up costs for customers or squeezing their own margins.
And this is turning pricing into one of the industry’s biggest competitive weapons. The companies that succeed in this new AI arms race are the ones that can deliver it at a price businesses are willing to pay.
