It has been found that ChatGPT-5 is faster, but not perfect
OpenAI has introduced the ChatGPT-5 model and declared it to be the fastest, smartest, and most reliable artificial intelligence with improved protection against hallucinations. This was reported by Zamin.uz.
However, practical research has shown that this new model is still not perfect. Despite the grand promises made in marketing statements, questions remain regarding its effectiveness and accuracy.
According to data from the Vectara platform regarding error frequency, ChatGPT-5 provides less incorrect information compared to GPT-4. Specifically, its error rate is 1.4 percent, while GPT-4 has an error rate of 1.8 percent.
Nevertheless, although ChatGPT-5 is slightly ahead of the GPT-4o model (1.4 percent compared to 1.49 percent), the difference is small, indicating strong competition. Interestingly, the ChatGPT-4.5 Preview version has an even better error rate of 1.2 percent.
The model with the least errors is o3-mini High Reasoning, which has an error rate of 0.79 percent. This indicates that other alternatives are seriously competing against ChatGPT-5's claim of being the "most reliable."
Despite improvements, ChatGPT-5 still lags behind the 4.5 model in terms of accuracy. At the same time, it significantly outperforms competitors like Gemini-2.5-pro (2.6 percent error) and Grok-4 (4.8 percent error).
Overall, ChatGPT-5 holds a leading position in the market, but it is not the absolute leader. After the model was launched, older GPT-4, 4o, and 4o-mini models were removed from OpenAI Plus subscribers without warning.
This led to dissatisfaction and debates among users. Many complained about the loss of traditional features, while some described it as "losing a single friend."
The sudden change in user habits and workflows naturally triggered emotional reactions. Technological changes must always be implemented with caution.
Company leader Sam Altman acknowledged that he did not sufficiently consider the importance of older versions for users and promised to temporarily reinstate the GPT-4o model. This decision was seen as an important step towards restoring trust.
The reinstatement is a sign of acknowledging shortcomings and engaging in open communication. Currently, there are two main directions.
On one hand, there are the achievements in speed, cognitive ability, and resistance to hallucinations of ChatGPT-5; on the other hand, the strong results of competing models like the 4.5 Preview and o3-mini High Reasoning are noticeable. The competition is very tight, and debates continue.
In conclusion, ChatGPT-5 can claim the title of "the fastest, smartest, and most reliable," but comparisons in the market also hold their place. Therefore, users are advised to try solutions that fit their workflows and needs.
If OpenAI remains true to its promises, results will speak louder than debates in the future.
