A mysterious AI model named ‘gpt2-chatbot’ recently appeared on the LMSYS Chatbot Arena. According to several users on X, the model showcased better reasoning and math abilities than OpenAI’s GPT-4. This left many users surprised, questioning whether it’s a new model by OpenAI. 

Interestingly, the model was released without official documentation, and there are no details to be found. However, soon after, OpenAI chief Sam Altman posted a cryptic message: ‘I do have a soft spot for GPT-2.’ This led to speculation that this could hint at a new version beyond GPT-4, possibly GPT-5.

AI influencer Rowan Cheung highlighted several notable features of the gpt2-chatbot, with its enhanced reasoning skills being praised by several users on X who posted screenshots.

A user on X  tested the chatbot’s mathematical capabilities, presenting it with an International Math Olympiad problem. Impressively, the chatbot solved it on the first attempt, although it couldn’t tackle all problems on the test. Despite this, its performance remained outstanding.

Moreover, the chatbot’s coding skills surpassed those of GPT-4 and Claude Opus, according to Chase, founding engineer at Codegen. He said the gpt2-chatbot excelled in complex code manipulation tasks, outperforming newer models. 

Furthermore, the chatbot’s proficiency in ASCII art was lauded by Cheung, who described it as “miles ahead of any other model” in this domain. 

Interestingly, this development comes as the tech ecosystem eagerly awaits GPT-5. Recently, Altman said that the company will release GPT-5 in the ‘coming months,’ adding that OpenAI has more important things to release before GPT-5. “Before we talk about a GPT -5-like model… I know we have a lot of other important things to release first,” said Altman. Meta also stirred the air with Llama 3, released about two weeks ago.