AI & Enterprise

OpenAI shared internal method to cut inference costs; could lower ChatGPT operating expenses

OpenAI engineers have reportedly shared internally an optimisation technique that can cut AI inference costs to less than half. Foreign media reports said the method was mentioned inside OpenAI in early June and has already been applied to some processing for ChatGPT guest users. Inference costs are ongoing expenses incurred whenever an AI model generates responses. Analysts have cited OpenAI’s cost burden, including an estimate that it spent more than $5 billion on inference in the first half of 2025.

AI & Enterprise

DeepSeek to double peak-hour fees for V4 model API from July

Chinese AI unicorn DeepSeek will double peak-hour usage fees for its V4 model API, shifting its pricing strategy after sparking a price war with deep discounts. The company told subscribers it will raise V4 API prices during peak hours of 9 a.m. to noon and 2 p.m. to 6 p.m. Beijing time. The changes will take effect when the full V4 version goes online in mid-July.