“GPT 4 was the dumbest model that any of us will ever have to use again” says Sam Altman

GPT-5 Emerges as OpenAI’s Most Capable Model Yet

OpenAI officially released GPT-5 on August 7, 2025, marking what the company describes as its most advanced and capable model so far. Built as a unified system that combines both high speed and deep reasoning modes, GPT-5 can dynamically switch between task-optimised variants behind the scenes. It supports text, image, voice, and video inputs, offers longer context handling, and can act as an autonomous agent for multi-step tasks.

It is the smartest version that OpenAI has put out till now: it has outperformed every existing model and is more useful to real-life queries now. It has a more compelling writing and, better response for health-related queries.

The model has been designed to excel in technical fields such as mathematics, health-related queries, real-world coding, and complex reasoning. OpenAI says it has also reduced hallucinations and implemented safer output generation.

Personality presets such as “Listener” and “Nerd” have been introduced to give users more control over interaction styles. Importantly, GPT-5 does not store any private information such as a user’s age, caste, or religion unless the user explicitly chooses to have it remembered.

Mixed Reactions as Users Miss the Old Warmth

Despite its improvements, GPT-5 has received a mixed reception from the community. Many long-time GPT-4 users say the new version feels colder and less emotionally engaging. On social media platforms like Reddit, people have compared the shift to “losing a friend or boyfriend” when older models such as GPT-4o were retired. Some have gone so far as to call the update a move toward a “corporate beige zombie” personality.

Frustrations have also surfaced over the removal of older versions. Users argue that GPT-4o’s combination of warmth and creativity is irreplaceable, and that they should have the freedom to choose which model they interact with.

GPT-5 has been criticised for hallucination, meaning it sometimes gives wrong answers. Users report that it can confidently provide inaccurate information or dismiss real questions. OpenAI is working to reduce hallucinations, but they remain a challenge for AI models like GPT-5.

Not all feedback has been critical. Some users and OpenAI staff point to its strengths in both collaboration and autonomy. “It’s a great coding collaborator, and also excels at agentic tasks,” says Michelle Pokrass, a post-training lead at OpenAI. “It executes long chains and tool calls effectively, follows detailed instructions, and provides upfront explanations of its actions. This means it better understands when and how to use functions like web browsers or external APIs.”

Some users appreciate GPT-5, saying it provides more straightforward and less wordy responses. Others believe that people had unrealistic expectations, reminding us that GPT-5 is still a large language model and not artificial general intelligence.

Ultimately, it is performing as intended.