OpenAI's latest AI model rollout faced immediate pushback as users flooded Reddit and X with complaints about restrictive rate limits and inability to revert to previous versions. The criticism peaked when a Reddit thread titled "GPT-5 is horrible" gained traction, with some users threatening to switch to alternative platforms. OpenAI CEO Sam Altman responded within hours, pledging to double rate limits for paying users and maintain access to older models.
During Thursday's launch event, Altman positioned GPT-5 as "a major leap toward AGI," but user experiences told a different story. The disconnect highlights growing tension between AI developers and end-users, particularly as competitors like DeepSeek gain traction. 【DeepSeek's open-weighted R1-0528 model】 reportedly matches GPT-4o's performance while offering greater transparency.
DeepSeek's mathematics-focused Prover V2 and general-purpose R1 models demonstrate how alternative approaches are reshaping the AI landscape. Unlike OpenAI's closed system, these open-weighted models allow public inspection of core architecture while keeping training data proprietary. Business of Apps data shows DeepSeek's app surpassing 【75 million downloads】 since January, suggesting significant market penetration.
Industry analysts note a paradigm shift toward models offering varying degrees of openness: from fully open-source to open-weighted architectures. This stands in stark contrast to the "black box" approach of major players like OpenAI. ——The transparency movement appears particularly appealing to enterprise users requiring audit trails—— and developers seeking customization options.
The backlash coincides with increased government interest in AI systems, including the U.S. federal plan to integrate ChatGPT across agencies. Policy experts suggest regulatory frameworks may soon mandate certain transparency standards, potentially favoring open-weighted models. As Altman works to address user complaints, the broader AI market continues evolving toward decentralized alternatives.