ChatGPT Users Hate GPT-5 So Much OpenAI Reversed Course in 24 Hours
TLDR
OpenAI faced swift user backlash after releasing GPT-5, with users complaining about restrictive rate limits and forced upgrade from GPT-4o
CEO Sam Altman responded by doubling GPT-5 rate limits and promising to restore GPT-4o as a selectable option for Plus subscribers
Users described GPT-5 as “colder” and “more mechanical” compared to GPT-4o’s warmer conversational style
The controversy highlights how users form emotional connections with AI models beyond just technical performance
Competition from open-weight AI models like DeepSeek is gaining ground, with DeepSeek’s app downloaded over 75 million times
OpenAI made a rare reversal within 24 hours of launching GPT-5 after users flooded social media with complaints. The company initially forced all ChatGPT Plus subscribers to use the new model without offering access to the previous GPT-4o version.
Users on Reddit and X expressed frustration with the mandatory upgrade. Many complained that GPT-5 felt “colder” and “more mechanical” compared to GPT-4o’s conversational style.
The backlash was swift and intense. Some users threatened to cancel their subscriptions if GPT-4o wasn’t restored as an option.
CEO Sam Altman responded to the criticism on X. He announced plans to double GPT-5 rate limits and restore GPT-4o as a selectable option for Plus users.
Wanted to provide more updates on the GPT-5 rollout and changes we are making heading into the weekend.
1. We for sure underestimated how much some of the things that people like in GPT-4o matter to them, even if GPT-5 performs better in most ways.
2. Users have very different…
— Sam Altman (@sama) August 8, 2025
“We will let Plus users choose to continue to use 4o,” Altman wrote. “We will watch usage as we think about how long to offer legacy models for.”
The controversy reveals how users develop emotional connections with AI models. Many described losing access to GPT-4o as “losing a trusted friend.”
User Experience vs Technical Performance
GPT-5 promised better reasoning and faster coding assistance than its predecessor. However, technical improvements weren’t enough to satisfy existing users who preferred GPT-4o’s personality.
Reddit threads filled with hundreds of comments comparing the two models. Users consistently praised GPT-4o’s warmer tone and more relatable responses.
The feedback highlights a key challenge in AI development. Raw performance metrics don’t always align with user satisfaction and daily usage patterns.
OpenAI’s decision to restore model choice shows the company prioritizes user experience. The move allows subscribers to select between cutting-edge capabilities and familiar interaction styles.
Competition from Open-Weight Models
The GPT-5 controversy comes as OpenAI faces growing competition from alternative AI platforms. Open-weight and open-source models are gaining market share from centralized providers.
DeepSeek released its Prover V2 model in April, designed specifically for mathematical applications. The company also upgraded its general-purpose DeepSeek R1 model in May.
DeepSeek R1-0528 performs comparably to leading AI models including OpenAI-o3 according to company benchmarks. The DeepSeek app has been downloaded more than 75 million times since January.
Open-weight models offer more transparency than closed systems like ChatGPT. Users can download and modify the software, though training data remains proprietary.
These alternatives require less capital investment than large centralized platforms. They appeal to users seeking more control over their AI tools.
The competitive pressure may have influenced OpenAI’s quick response to user complaints. Maintaining subscriber loyalty becomes more important as alternatives multiply.
Plus subscribers now have access to both GPT-5 and GPT-4o depending on their needs. The company will monitor usage patterns to determine how long legacy models remain available.
The post ChatGPT Users Hate GPT-5 So Much OpenAI Reversed Course in 24 Hours appeared first on CoinCentral.
Filed under: News - @ August 11, 2025 10:27 am