Newsletters
OpenAI has reduced prices on its lower-tier and mid-tier artificial intelligence models as businesses face mounting pressure over rising AI costs, according to Reuters.
Tokens, the basic unit used to measure AI usage, refer to the pieces of text processed when a model reads or generates content.
The company cut prices for its smaller GPT-5.6 Luna model by 80% and for its mid-tier Terra model by 20%. Pricing for Sol, OpenAI's flagship model, remains unchanged.
The reductions increase competitive pressure on Anthropic, whose Claude models are widely used among enterprise clients and developers.
Anthropic's mid-tier Claude Sonnet 4.6 is priced at $3 per million input tokens and $15 per million output tokens, above Terra's revised rates.
Both companies face pressure from lower-cost, open-source Chinese competitors such as Z.ai's GLM-5.2 model, which offers comparable performance at a reduced cost.
Under the revised pricing, Luna's text input rate dropped from $1 to $0.20 per million tokens, while its output rate fell from $6 to $1.20. Terra's input rate decreased from $2.50 to $2, and its output rate dropped from $15 to $12.
OpenAI said part of the price reduction was made possible by efficiency gains from GPT-5.6, citing the model's ability to optimize performance in code development and internal processes.
The cuts, announced Thursday, apply only to OpenAI's small and mid-tier models. The company said these models can now perform tasks that previously required higher-tier systems, at a significantly lower cost, which it said benefits businesses overall.
Analysts say the price cuts could expand adoption of OpenAI and Anthropic's technology but may strain the companies' finances ahead of long-anticipated public offerings.
While token prices have declined industry-wide over the past year, the shift from fixed subscription models to usage-based pricing has increased the total cost of specific tasks, making per-task spending harder to predict and, in many cases, higher for businesses.
Set as preferred source
Set as preferred source
2 min read
