GPT-4.5 Turbo模型疑似提前泄露的消息近日在网络上引起广泛关注。据网友发现，该模型已在Bing、谷歌和DuckDuck Go等搜索引擎中出现，并被描述为OpenAI迄今为止最快速、准确且可扩展性最强的模型。其上下文窗口高达256K，是GPT-4-Turbo的两倍，预计可处理约200,000个单词。
据推测，GPT-4.5 Turbo的官方发布会可能选在本周四，即GPT-4发布一周年之际。这一推测得到了OpenAI首席执行官Sam Altman即将参加Lex Fridman播客节目的佐证。去年，Altman在GPT-4发布后参加了Fridman的播客。
此外，泄露信息还显示，GPT-4.5 Turbo的“知识更新截止日期”为2024年6月，暗示原定于6月发布的模型可能已被提前。
英文翻译内容：
Title: GPT-4.5 Turbo Leak Sparks Speculation of Thursday Release
Keywords: AI Model Leak, GPT-4.5 Turbo, OpenAI
News content:
Recent online chatter has focused on the potential early release of the GPT-4.5 Turbo AI model, with users finding references to it across search engines like Bing, Google, and DuckDuck Go. Described as the fastest, most accurate, and scalable model from OpenAI to date, GPT-4.5 Turbo boasts a 256K context window, double that of GPT-4-Turbo, capable of processing approximately 200,000 words.
Speculation is rife that the official launch of GPT-4.5 Turbo could occur this Thursday, marking the one-year anniversary of GPT-4’s release. This theory gains traction from the fact that OpenAI CEO Sam Altman is set to appear on Lex Fridman’s podcast, following a similar appearance last year just after GPT-4 was announced.
Additionally, the leaked information indicates that the “knowledge cutoff date” for GPT-4.5 Turbo is June 2024, suggesting that the model, originally slated for a June release, may have been significantly advanced.
【来源】https://mp.weixin.qq.com/s/hfOClnutRNguxmFdjkggvA
Views: 1