OpenAI cuts prices of small AI models
OpenAI lowered prices for its low- and mid-tier artificial intelligence models on Thursday, a move that could intensify competition in the sector as U.S. firms compete with Chinese rivals offering cheaper alternatives to attract customers increasingly concerned about the high costs of using the technology.
The company, developer of the chatbot ChatGPT, cut the cost of its smallest model (GPT-5.6 Luna) by 80% and its mid-tier model (Terra) by 20%, while keeping the price of its largest and flagship model (Sol) unchanged.
These reductions indicate that increased scrutiny from companies facing high AI usage bills is forcing U.S. AI labs to reconsider their pricing policies, according to Reuters.
Several technology CEOs have recently emphasized that providing more affordable options is a key factor in the widespread adoption of AI.
OpenAI’s new pricing policy increases pressure on Anthropic, whose Claude models dominate corporate and developer usage but remain among the most expensive in the market.
Both companies face pressure from Chinese competitors offering open-source models, such as Z.ai’s GLM-5.2, which approaches their performance levels at a lower cost.
Analysts said the price cuts could boost the use of OpenAI and Anthropic technologies, but may strain their financial positions ahead of the initial public offerings anticipated by investors.
Although the Thursday price reductions apply only to OpenAI’s small and mid-tier models, the company stated that businesses will benefit broadly, as these models can now perform tasks that previously required higher-tier models, but at a significantly lower cost.
Under the new pricing, companies using OpenAI technologies will pay less per million “tokens,” the unit of measurement used to calculate AI model usage.
The cost of sending text to the Luna model dropped to 20 cents per million tokens, down from $1, while the Terra model’s cost fell to $2 from $2.50. Additionally, the cost of generating responses decreased to $1.20 and $12, down from $6 and $15, respectively.
In comparison, Anthropic’s mid-tier Claude Sonnet 4.6 model costs $3 per million input tokens and $15 per million output tokens, prices higher than those of the Terra model.
OpenAI said the price reductions were made possible in part by efficiency gains achieved by the GPT-5.6 model, including its ability to optimize coding and improve performance efficiency during internal development processes.
Overall, token prices have declined over the past year, but the cost of completing tasks is rising as AI companies shift from fixed subscriptions to usage-based pricing.
This shift leads to unexpected, often higher bills, due to the difficulty of estimating usage per task.
