OpenAI’s release of GPT-5 on August 7, 2025 was intended to mark a major leap forward, but user reaction quickly turned sour. Thousands of posts on Reddit and other forums voiced frustration, with many users deeming the new model a downgrade despite CEO Sam Altman’s ambitious launch presentation.

In one Reddit thread aptly titled “GPT-5 is horrible,” complaints poured in. Users described responses as brief, bland, or inconsistent. One user criticized the company’s promotion of GPT-5, arguing the demo offered poor advice on handling mistakes and lacked the depth or nuance of earlier conversations.

Paid subscribers felt particularly slighted. Long gone are the variety of GPT-4o mini and high-mini options. Now users must contend with “GPT-5 Thinking,” which is capped at 200 messages per week. This limitation was seen as an unwelcome restriction rather than improvement.

Critics also lamented the automatic routing system that replaced older models entirely with GPT-5. Many users wrote that they missed the familiarity and warmth of GPT-4o and GPT-4.1. Comments like “I miss 4.1. Bring it back” were common, signaling emotional attachment to the previous model.

Part of the backlash stemmed from the hype preceding GPT-5’s reveal. Hours before the unveiling, Altman posted an image of the Death Star from Star Wars, implying a groundbreaking shift. Instead, many users felt the update failed to deliver relatable or emotionally engaging interactions.

Despite strong benchmark results, GPT-5’s real-world performance is being called into question. Users argue the model feels strained compared to GPT-4o and see GPT-5’s limitations—both in tone and accessibility—as undermining the value of their subscriptions.

OpenAI responded by reinstating GPT-4o for paid users, at least temporarily. The company acknowledged how users grew attached to that version’s personality and said future updates would allow more customization and steerability within GPT-5.