OpenAI launched GPT-5, sparking a mix of praise and criticism within the community over the past few days. This article breaks down the initial user feedback. (Background: Comprehensive summary of OpenAI's powerful GPT-5 launch: Free access, feature highlights, gpt-5, gpt-5-mini, and gpt-5-nano API pricing) (Additional context: OpenAI's open-source weight model GPT-OSS: How does Apache 2.0 crown developers? CoT thinking chain unlocks security sovereignty) This Friday morning (8th), OpenAI officially released the latest model GPT-5, along with sub-models like 'thinking', 'mini', and 'nano', making them available for both free and paid users. Automatic model selection: Convenience or black box? The community feedback over the past few days has raised concerns, with some claiming it has become smarter, while others complain that under the automatic selection of new models, GPT-5 has assigned them to weaker versions, and many users have posted cases of 'Nano producing incorrect answers'. Currently, OpenAI's new mechanism has eliminated manual model selection, forcing users to accept the system's judgment. For general users, this reduces selection anxiety; however, advanced users feel uncertain whether the response is from GPT-5 thinking or Nano, making it hard to adjust their prompting strategies. Senior AI observers point out that this 'one-click usage' simplified design is friendly to the public. After all, if every simple question is answered by the most powerful model, it is indeed not cost-effective, and speed may slow down; however, it could undermine the sense of control for professional users. Pricing and permissions: The outcry of Plus users The pricing of GPT-5 has also sparked discussion. Some users complain that the Plus plan has been 'severely weakened', and only those with the $200 Pro subscription enjoy all the features. They criticize: 'ChatGPT has become worse for Plus users, the automatic selection almost always sends you to the lower-tier model.' However, some users maintain a positive evaluation, with one user intentionally adding a small mistake when analyzing corporate financial reports with GPT-5, resulting in all models catching the contradiction; in contrast, GPT-4.1 and the old Mini mostly judged 'no logical error'. Additionally, some have summarized that GPT-5 is 'more willing to provide clear decision-making suggestions' compared to its predecessor, reducing the avoidance of answers that used to hedge bets on both sides. As for the tone of the responses, some believe GPT-5 is more neutral, making it more suitable as an assistant; while some prefer the more flattering tone of 4o, which may be subjective. Altman in the spotlight In response to online criticism that GPT-5 is not significantly better and a preference for 4o, Sam Altman responded yesterday that Plus users will be allowed to use 4o again, and provided the following responses: GPT-5 rollout updates: We will double the usage limits for ChatGPT Plus users as we complete the rollout. Plus users can choose to continue using 4o. We will monitor usage and evaluate how long to continue offering the legacy models. Starting today, GPT-5 will appear smarter. Yesterday, the automatic switch was down for a while, causing GPT-5 to perform poorly. Additionally, we are adjusting the decision boundary judgment method to ensure you receive the correct model more frequently. We will make information about the response model more transparent, allowing you to clearly know which model is answering your question. The UI interface will be updated for easier manual triggering of deep thinking mode. The overall rollout progress is slightly slower than expected. This is a massive change; for instance, in the past 24 hours, our API traffic has nearly doubled. We will continue to work on stabilizing the system and will keep listening to user feedback. As mentioned earlier, we originally anticipated that pushing out so many changes at once would be somewhat bumpy, but… this has been bumpier than we expected! GPT-5 rollout updates: *We are going to double GPT-5 rate limits for ChatGPT Plus users as we finish rollout. *We will let Plus users choose to continue to use 4o. We will watch usage as we think about how long to offer legacy models for. *GPT-5 will seem smarter starting… — Sam Altman (@sama) August 8, 2025 Next steps to observe Currently, GPT-5 has only been online for three days, and users need more time for testing. Whether OpenAI can achieve a balance between stabilizing quality, transparent model allocation, and reasonable fees will determine if GPT-5 can truly establish itself in the next competitive stage. Related reports Is your TikTok Shop still safe? AI + fake apps + phishing pages, the new type of scam 'ClickTok' is targeting your crypto wallet How long has it been since you seriously thought? Princeton scholar warns: AI is eroding the 'heritage of the Enlightenment' (Lawyer Lin Shanglun's article) Did AI defeat three lawyers? Don't get it wrong, this is the prologue of 'Lawyer 2.0' "Useful or disappointing? GPT-5 hands-on experience summary: A community debate of 'praise' and 'criticism' coexist" This article was originally published on BlockTempo (the most influential blockchain news media).