GPT 4 vs GPT 5: Here’s what’s improved and what’s not so great with ChatGPT's latest update

OpenAI's swelled linguistic communication pose update was eventually revealed to the public on th and soon started rolling out to all users. After the GPT-5 launch, OpenAI decided to discontinue its older models running ChatGPT like GPT-4o, GPT-4.1, o3, o3 Pro, o4-mini and more.

For most use cases, GPT-5 takes over the reins from GPT-4o, the large language model which replaced GPT-4 as the default model on ChatGPT earlier this year.

In case you are wondering what exactly has changed in ChatGPT with the new model, let's take a detailed look.

GPT-5 vs GPT-4o: What's new

1) Technical changes:

OpenAI says that GPT-5 marks a big leap in intelligence over all its previous models, with the new model said to boast state-of-the-art performance across coding, math, writing, health and visual perception.

OpenAI CEO Sam Altman even went on to tout GPT-5 as a “PhD-level expert” across various domains.

2) Faster responses and less hallucination:

OpenAI also claims that GPT-5 is much faster and more accurate than its previous models. Meanwhile, the AI startup states that GPT-5 is less likely to hallucinate (make stuff up) compared to its previous models.

GPT-5 responses were 45% less likely to contain a factual error than GPT-4o and 80% less likely than o3 with web search enabled on anonymized prompts, OpenAI stated.

It's worth noting that while OpenAI has focused on speed and accuracy a lot with GPT-5, many users on social media have claimed to receive slower and inaccurate responses compared to GPT-4o.

3) Better coding:

OpenAI says GPT-5 is its strongest model for coding to date, with improvements in complex front-end generation and debugging larger repositories. ChatGPT running GPT-5 can now create responsive websites, apps and games with just simple prompts in natural language, reducing the need for users to have any coding abilities.

4) Better writing:

GPT-5 is also said to be the ‘most capable writing collaborator’ to date by OpenAI, which the company says can translate rough ideas into compelling, resonant writing with literary depth and rhythm.

The social media reviews for GPT-5 for writing, however, seem to be varied, with some users reporting better narrative consistency and improved rhythm and flow, while others state that the new model is less creative than GPT-4o and struggles with character consistency while giving shorter responses with less emotional depth.

5) Health related questions:

OpenAI has made big improvements with GPT-5 in answering health related questions, which the company says will empower users to be more informed and advocate for their health. The company states that GPT-5 acts more like an active thought partner by proactively flagging potential concerns and asking questions to give answers to their health-related queries.

"The model also now provides more precise and reliable responses, adapting to the user’s context, knowledge level, and geography, enabling it to provide safer and more helpful responses in a wide range of scenarios," OpenAI said in its blog post.

6) Unified interface:

OpenAI has officially removed the model picker from ChatGPT, meaning that users can no longer choose which model they want to interact with. GPT-5 instead contains an efficiency model and a reasoning model, but switching between them isn't as easy as earlier, with a real-time router now deciding which queries are sent to the thinking model based on the conversation type, complexity, tool needs, and query intent.

7) Usage limits:

At launch, the GPT-5 usage limits were more or less unchanged for all users, with free-tier users having the ability to send 10 messages every 5 hours, Plus users having 80 messages every 3 hours, and Pro users getting unlimited messages for the standard version.

However, with OpenAI facing a lot of flak for removing the older GPT models, the company has decided to temporarily double the limit for Plus users to 160 messages every 3 hours.

Once the usage limit is breached, other queries of the user are then handled by a Mini version.

8) Reduced personality:

OpenAI has managed to reduce sycophancy (being too agreeable with users) on ChatGPT with the GPT-5 update, but long-time users are now complaining that the latest model doesn't have much personality compared to the incumbent ones. Many users who had grown accustomed to using ChatGPT as their friend, therapist, or just a companion of sorts state that the new model lacks the emotional depth of GPT-4o and feels sterile, formal, and less engaging.

Sam Altman has also commented on the matter and stated that after the GPT-5 rollout to all users, OpenAI would work on making the new model “warmer.”

---

---