> Be OpenAI > GPT-3 to GPT-4 was peak kino. > GPT-5 will show real improvements over its predecessors, but they won’t be comparable to leaps in performance > OAI can’t clearly outpace rivals anymore > 73% of OAI revenue comes from chatbot subscriptions >“700 million weekly” https://t.co/kojvW7Hdke
OPENAI CHATGPT PAID BUSINESS USERS GREW 66% IN JULY MONTH OVER MONTH TO 5M USERS This was fueled by larger deployments and POC conversions. https://t.co/iVBlqytHoE
关于 GPT-4.5/5 及更多信息的摘要:太长不看版 GPT-4.5 (“Orion” / 猎户座) * 最初以 Orion 为代号开发,并计划作为 GPT-5 发布。 * 性能令人失望:与 GPT-4o 相比没有重大飞跃。 * 失败原因: * 用于预训练的高质量网络数据日益枯竭。 * 优化方法在小模型上有效,但无法扩展到大模型。 * https://t.co/K9FWtseyr8
OpenAI’s chief operating officer said the company’s ChatGPT platform now serves more than 5 million paying business users, a figure that grew 66% in July alone as large-scale deployments and proof-of-concept conversions accelerated. The New York Times has reported that subscriptions to the chatbot generate about 73% of OpenAI’s revenue, underscoring the importance of enterprise uptake to the privately held firm’s business model. The surge in customers comes as OpenAI wrestles with technical headwinds in developing its next-generation model, GPT-5. According to an article by The Information, an interim system dubbed GPT-4.5, or “Orion,” was originally slated to carry the GPT-5 label but was rebranded after failing to provide a breakthrough comparable to the leap from GPT-3 to GPT-4. Researchers cited a shrinking supply of high-quality training data and difficulty scaling optimization methods, suggesting that industry-wide progress in large language models may be slowing even as commercial demand rises.