gpt-5 是gpt-4的后续产品，gpt-4是openai ⽬前最先进的ai 聊天机器⼈，您需要付费进⾏订阅才能使⽤。 GPT 代表⽣成式预训练变压器，它是⼀种⼤型语⾔模型，可以创建类似⼈类的⽂字和图像等内容。

2024年3月24日 · Overall performance boost: Sam Altman, CEO, stated that GPT-5 will be smarter and superior in all aspects, with a significant performance improvement over GPT-4. Enhanced multimodal capabilities: It is predicted that GPT-5 will not only handle text and images but will also be capable of processing audio and videos, becoming a multimodal AI.

Jimmy Apples said OpenAI finished training a 125T parameter GPT-5 model in October (He could have assumed it was GPT-5, i know it as Arrakis). I mean i thought it might have finished in December, but October was a bit of a suprise. Very capable model. Internally OpenAI's GPT-5 goal was ~100x GPT-4, and well they have done that.

2024年3月20日 · GPT-4.5 or 5 is apparently estimated to be capable of automating around 3% of the worlds jobs according to sources that are pretty close to Sam Altman. I’m predicting GPT-4.5 will be the first release of a new model + new interface that has autonomous agent capabilities, possibly with the AI having it’s own virtual environment to navigate ...

今天new gpt-5神不知鬼不觉的登场了，好多人还不知道。我先教方法。 你必须是gpt-plus用户，也就是有付费每月20刀的那个。 在谷歌中搜索new gpt-5，然后第一个结果你会看见官网里面多了一个new gpt的连接。 点击进去，即可开始享受new gpt-5

A year ago, we trained GPT-3.5 as a first “test run” of the system. We found and fixed some bugs and improved our theoretical foundations. As a result, our GPT-4 training run was (for us at least!) unprecedentedly stable, becoming our first large model whose training performance we were able to accurately predict ahead of time.

2024年9月1日 · Even Sam Altman said that they were considering releasing an early checkpoint of GPT 5's training run and calling it GPT 4.5 or something else because they want to avoid "big releases." I wouldn't be surprised at all to learn that GPT 5 is still training and what these people had access to was a checkpoint.

2023年11月17日 · "GPT-4" is the name of the model and architecture, rather than the features it has. It's a bit like having a child called GPT-4 and they gradually get new capabilities. That child can learn new things and be given new tools but it will still be GPT-4. It's not until you have a new child trained from scratch that you'd say it's GPT-5.

2024年4月30日 · The model could be an early version of GPT-5 or a heavily modified GPT-2, speculated because of its advanced capabilities and use of OpenAI's tiktoken tokenizer. It solved an International Math Olympiad problem in one …

2024年3月12日 · I mean even logically he could just be talking about both models relative to GPT-3/3.5, in which case it's an obvious truth just on the basis that GPT-5 is a linear improvement. I don't think that's the case, but it illustrates how the statement can't really be called out as a lie with any reasonable veracity. It's perfectly acceptable marketing.