OpenAI has significantly cut prices for its latest GPT-5.6 models in an effort to maintain its competitive edge against emerging Chinese artificial intelligence providers. The US-based company announced on Thursday price reductions of up to 80% for its lightweight GPT-5.6 Luna model, reducing application programming interface (API) fees to $0.20 per million input tokens and $1.20 per million output tokens. These fees represent the cost developers pay to embed AI functionality in their applications.

Alongside the steep discount for the Luna model, OpenAI lowered prices for its mid-tier GPT-5.6 Terra model by 20%, setting rates at $2 per million input tokens and $12 per million output tokens. The company also unveiled a new "fast mode" for its flagship GPT-5.6 Sol model, which it said would boost processing speeds by up to 2.5 times. This suite of models was initially launched last month.

The price adjustments have propelled GPT-5.6 Luna to the top of intelligence-per-dollar rankings, according to the research firm Artificial Analysis. This marks a competitive advantage over Chinese rivals such as Zhipu AI’s GLM-5.2 and MiniMax’s M3 models.

The move comes amid intensifying competition as Chinese AI companies gain traction with open-weight architectures and aggressive pricing strategies. Earlier this week, Moonshot AI, headquartered in Beijing, released API access to its KimK 3 model, which outperformed OpenAI’s GPT-5.6 Sol and Anthropic’s ClaudeFable 5 in benchmark tests focused on frontend development tasks. On the Artificial Analysis Intelligence Index, KimK 3 scored above OpenAI’s Terra and Luna models but still lagged behind Sol.

However, Moonshot AI recently increased KimK 3’s commercial pricing to $3 per million input tokens and $15 per million output tokens, making it more expensive than OpenAI’s revised Terra and Luna tiers.

Investment bank Jefferies highlighted cost-efficiency as a crucial factor in this evolving AI market. While Chinese models have traditionally been priced significantly lower than their Western counterparts, recent price cuts by major US firms have narrowed this gap.

Despite OpenAI’s reductions, several Chinese models remain competitively priced. Zhipu AI’s GLM-5.2 charges $1.40 per million input tokens and $4.40 per million output tokens, while MiniMax’s M3 is priced as low as $0.30 per million input tokens and $0 for output tokens. DeepSeek’s V4-Pro also offers affordable rates at $0.45 per million input tokens and $0.87 per million output tokens.

The latest pricing strategies underscore the increasing emphasis on value-for-money metrics as AI developers worldwide weigh cost against performance in deploying advanced language models.