① OpenAI announced on Thursday a significant price reduction for its two latest AI models, cutting the price of GPT-5.6 Luna by 80% and that of GPT-5.6 Terra by 20%; ② enterprise customers are becoming increasingly cost-sensitive, and OpenAI must also fend off competition from multiple tech giants, including Google and Microsoft.
Caixin Global, July 31 (Editor Bian Chun) — OpenAI announced on Thursday a substantial price cut for its two newest artificial intelligence (AI) models, GPT-5.6 Terra and GPT-5.6 Luna, just about three weeks after their public launch.
Enterprise customers are growing more sensitive to costs and are reluctant to deploy expensive models unless they can clearly see a return on investment. OpenAI is under pressure to meet these customer demands while simultaneously defending its position against intense competition from tech giants such as Google and Microsoft, who are aggressively promoting high-performance, cost-effective models.
On July 9, OpenAI officially launched its new model series—GPT-5.6 Sol, Terra, and Luna. The Sol version serves as the flagship model of the GPT-5.6 series, with OpenAI highlighting its enhanced capabilities in programming, biology, and cybersecurity; the Terra version is designed for everyday professional tasks; and the Luna version offers speed and affordability.
On Thursday, OpenAI stated it would reduce the price of GPT-5.6 Luna by 80% and that of GPT-5.6 Terra by 20%. The Luna API is now priced at $0.20 per million input tokens and $1.20 per million output tokens; the Terra API is priced at $2 per million input tokens and $12 per million output tokens.
In addition, pricing for SOL, ChatGPT, and Codex subscriptions, as well as quota budgets, remains unchanged; Terra and Luna now consume fewer credits.
“Our core strategy remains to simultaneously enhance model capabilities and operational efficiency, striving to enable each new generation of intelligent models to accomplish more tasks at lower costs,” OpenAI said in its statement.
The 'token price war' intensifies
The launch of ChatGPT in 2022 sparked an AI boom, prompting U.S. companies to rush to deploy the technology and encouraging employees to use it extensively. This gave rise to the so-called era of 'token maxxing,' during which employers urged staff to use AI as much as possible without concern for cost.
However, as AI-related bills have soared—with some corporate expenditures reaching billions of dollars—many companies have begun actively managing their spending. This shift has drawn significant attention from AI developers. Meanwhile, a wave of open-weight models has rapidly emerged, with performance steadily approaching that of leading closed-source models.
Open-source AI models allow users to download, modify, and run them on their own infrastructure, offering enterprises a cost-effective option. Earlier this month, Chinese startup Moonshot AI released an open-source model named Kimi K3, which performs comparably to leading U.S. models on certain tasks, prompting a swift response from Silicon Valley.
Following Moonshot AI's release of Kimi K3, Anthropic—OpenAI’s top rival—launched a new model, Claude Opus 5, touting it as the best-performing and most cost-effective option across a wide range of applications. Although Claude Opus 5 delivers performance on par with Claude Fable 5, Anthropic’s high-end model released in June this year, particularly in programming and knowledge-based tasks, it is priced at only half the cost.
Microsoft CEO Satya Nadella also repeatedly emphasized the company’s cost-effective models during the quarterly earnings call on Wednesday. Earlier this week, Microsoft unveiled a cybersecurity model it described as 'inexpensive yet high-performing.'
Google also introduced three new models this month, designed to outcompete rivals on cost. The company further stated that its highest-performing newly launched model, Gemini 3.6 Flash, has a lower per-task cost than Kimi K3.
Editor/KOKO