OpenAI has slashed prices for two of its newest artificial intelligence (AI) models, escalating a fierce price war as tech giants and open-source rivals battle to win over enterprise customers growing increasingly sensitive to ballooning AI bills.
Company CEO Sam Altman announced the price cuts on Thursday, touting cheaper access to GPT-5.6 Terra and GPT-5.6 Luna just three weeks after their rollout. Under the new scheme, the mid-tier Terra model will see a 20% price reduction to $2 per million input tokens and $12 per million output tokens. The high-speed Luna model will drop by 80%, falling to 20 cents per million input tokens and $1.20 per million output tokens.
“We want to offer the best price/intelligence tradeoff at every level,” Altman wrote on X.
Pricing for OpenAI’s top-tier reasoning model, GPT-5.6 Sol, remains unchanged at $5 per million input tokens and $30 per million output tokens, though the company introduced a higher-throughput Sol Fast tier for double the price. OpenAI noted that the cost savings will extend to paid enterprise subscriptions, including Codex and ChatGPT Work.
The dramatic price cuts underscore a broader shift in the AI market. After years of unchecked spending —an era analysts dubbed tokenmaxxing — enterprises are demanding a clearer return on investment.
“Enterprises have figured out how easy it is to burn tokens without getting value back, and they’re pushing back on those increasing AI bills,” said Jacob Bourne, senior analyst at eMarketer.
OpenAI attributed its lower rates to structural efficiencies across its software stack, including improved model routing, optimized production code, and smarter agentic harnesses that prevent AI agents from repeating tasks.
The move comes amid intense pressure from both domestic competitors and international developers. Chinese startup Moonshot AI recently rattled Silicon Valley by releasing its open-weight Kimi K3 model, which offers high-end performance at a fraction of the cost of proprietary Western models.
Rival labs have responded swiftly: Anthropic recently unveiled Claude Opus 5 at half the cost of its previous flagship, while Google debuted Gemini 3.6 Flash and Gemini 3.5 Flash-Lite to target low-cost inference workloads.
Microsoft Corp. CEO Satya Nadella also emphasized “cost efficiency” during a recent earnings call, highlighting the tech giant’s new MAI-Thinking-1 model architecture designed to lower overall task costs.
The strategic pivot to flexible, low-cost pricing comes at a critical juncture for OpenAI, which recently filed confidentially for an initial public offering.
Arun Chandrasekaran, a distinguished vice president analyst at Gartner, noted that the move marks a transition away from rigid commercial structures toward flexible pricing, serving as an early test for frontier AI labs ahead of anticipated public listings.