OpenAI significantly reduces prices! GPT-5.6 Luna drops by 80%: cost-effectiveness surpasses DeepSeek V4 Pro

OpenAI officially announced the new pricing plan for the GPT-5.6 series. This price adjustment only applies to the Terra and Luna mid-range and lightweight models, while the flagship GPT-5.6 Sol pricing remains unchanged.

Among them, Luna, designed for batch lightweight reasoning scenarios, has seen the largest price cut in history, with both input and output costs reduced by 80%.

Before the adjustment, GPT-5.6 Luna was priced at $1 per million input tokens (approximately RMB 6.8) and $6 per million output tokens (approximately RMB 40.8). After the adjustment, input is only $0.2 per million tokens (approximately RMB 1.4), and output is $1.2 per million tokens (approximately RMB 8.1), significantly reducing usage costs.

The balanced version Terra in the same series has been reduced by 20%, with input at $2 and output at $12 per million tokens, catering to moderate complex reasoning needs. Meanwhile, Sol, which focuses on complex code and multimodal deep thinking, maintains its original high price, forming a clear tiered pricing strategy: low-cost Luna seizes the batch reasoning market, while high-end Sol secures the high-margin business base.

OpenAI also released evaluation charts from the third-party organization Artificial Analysis Intelligence Index v4.1, comparing mainstream lightweight commercial models in terms of unit task cost and overall model score. The data shows that the newly priced GPT-5.6 Luna comprehensively outperforms DeepSeek V4 Pro in overall cost-effectiveness for completing equivalent standard tasks.

In general scenarios such as text generation, multi-turn conversations, and tool calls, it achieves higher overall scores and lower single-task comprehensive costs, addressing the previous cost disadvantage of overseas models compared to cost-effective domestic products.

Industry analysis suggests that this significant price reduction is primarily driven by the sustained low-price impact of domestic large models. DeepSeek V4 Pro leverages a caching mechanism to achieve extremely low recurring input costs, with regular reasoning pricing remaining low during off-peak hours. Coupled with multiple domestic manufacturers continuously offering affordable API services, this has consistently diverted overseas small and medium-sized developer customers, forcing OpenAI to lower pricing for entry-level models.