GPT-5 vs GPT-4o: Is the latest OpenAI model better than its most beloved one?

The most recent launch of GPT-5 has taken the world by storm. OpenAI’s latest flagship mannequin has acquired blended critiques – whereas some reward its capabilities, others spotlight its shortcomings. This made me marvel: Is GPT-5 really superior to the unique favourite, GPT-4o?

Personally, GPT-4o was my go-to LLM for all the things from textual content summarization to picture technology and information evaluation. Now that OpenAI has changed it with GPT-5, I made a decision to place each fashions to the check. Is that this improve genuinely evolutionary, or a rushed transfer which may diminish ChatGPT’s attraction?

Let the battle of the GPTs start!

GPT 5 and GPT 4o: A Fast Reminder

Let’s  shortly dive into particulars concerning the two chatGPT fashions that we’ll be testing on this weblog: GPT-5 and GPT 4o

GPT-5

Launched final week, GPT-5 now stands as ChatGPT’s most superior mannequin. OpenAI’s newest multimodal LLM introduces agentic capabilities and a ‘unified system’ for job evaluation. This method mechanically determines whether or not a question requires deep reasoning or primary processing. In contrast to earlier fashions, GPT-5 follows a ‘learn-by-doing’ method. It reveals elevated empathy whereas being much less agreeable than its predecessors. Together with this GPT-5 comes with enhanced coding, writing and vibecoding powers

Discover extra in my earlier article on GPT-5.

GPT-4o

Launched final yr, GPT-4o (the place “o” means “omni”) was the first-of-its-kind mannequin. This multimodal modified the way in which individuals used ChatGPT. The mannequin got here with enhanced coding and visible evaluation capabilities. GPT-4o got here with speech recognition and speech evaluation options too. The mannequin got here with elevated processing pace and diminished response latency. OpenAI’s GPT-4o generated extra pure and smart responses, and was in a position to entry instruments and provides real-time data.

To know extra, checkout this text on GPT 4o.

GPT 5 vs GPT 4o: Characteristic Comparability

GPT 5 vs GPT 4o: Process Comparability

Now, let’s put each fashions to the check by evaluating their efficiency on the next duties:

1. Content material creation
2. Picture Era
3. Coding
4. Picture Evaluation
5. Reasoning

Let the GPT-5 vs GPT-4o battle start!

Process 1: Content material Creation

Learn the article at https://www.analyticsvidhya.com/weblog/2024/07/building-agentic-rag-systems-with-langgraph/ to know the method of making a vector database for Wikipedia information. Then, present a concise abstract of the important thing steps.”

GPT-5 Response:

GPT-4o Response:

Commentary:

The response generated by GPT-5 is the concise abstract that an individual educated on the subject would need. The steps are all listed in correct order and include simply sufficient context. However, GPT-4o’s response is a abstract of all of the steps talked about within the weblog. It lists all of the steps that have been coated within the weblog in the identical method as mentioned there. The distinction within the method of the 2 fashions is: GPT-5 merges the factors to generate a concise abstract of the whole course of, whereas GPT-4o creates a concise abstract of all of the steps coated within the weblog.

Process 2: Picture Era

The picture is of working of a voice agent. It has 3 predominant components
Speech-to-text (STT): Captures and converts your spoken phrases into textual content.
Agentic logic: That is your code (or your agent), which figures out the suitable response
Textual content-to-speech (TTS): Converts the agent’s textual content reply again into audio that’s spoken aloud.
Convert this primary picture into vibrant picture.

GPT-5 Response:

GPT-4o Response:

Commentary:

The duty was easy, and each fashions executed it fairly effectively. Beginning with GPT-5, it created a vibrant picture with popping colours. The picture it generated had textual content and icons; nevertheless, there was a minor error – a small arrow connecting the mic icon with the TTS field. As for the picture generated by GPT-4o, it used strong colours, making it much less vibrant. The strengths of GPT-4o’s picture have been the audio enter and output sources that it included.

Process 3: Coding

Primary HTML code for a word-counting web site.

GPT-5 Response:

GPT-4o Response:

Commentary:

GPT-5 took a while to generate the code for this question, particularly for the phrase counter web site. Nonetheless, the ultimate output was fairly spectacular. The UI/UX and options got here collectively to create a totally practical word-counting webpage. However, GPT-4o’s output felt lackluster compared. The UI/UX was primary, providing solely the core word-counting characteristic with out further refinements. Its design additionally appeared considerably outdated

Process 4: Picture Evaluation

Calculate the output of this circuit diagram.

GPT-5 Response:

GPT-4o Response:

Commentary:

GPT-5 answered this query shortly, analyzing each the picture and its parts effectively. It accurately recognized the half-wave rectifier, learn the values marked on the diagram, and utilized the correct logic to calculate the output present and voltage values. In distinction, GPT-4o struggled with this job. Whereas it acknowledged the output waveform, it did not course of different vital elements. Most notably, GPT-4o couldn’t extract the mandatory values from the picture to carry out any calculations.

Process 5: Reasoning

Clear up the next Sudoku and provides the ultimate answer as a picture.

GPT-5 Response:

GPT-4o Response:

Commentary:

GPT-5 initially struggled with picture interpretation, taking up three minutes to course of the enter. Reasonably than fixing the puzzle independently, it requested affirmation of a number of values throughout the picture. After I manually offered all of the row values, the mannequin efficiently processed and solved the puzzle, yielding an accurate answer, although requiring important person help.

GPT-4o, in contrast, failed to resolve the puzzle totally. It merely populated all lacking values with zeros and offered this as its output answer.

GPT-5 vs GPT-4o: Closing Verdict

Deciding on a transparent winner has by no means been more difficult. Right here’s how the 2 LLMs carried out throughout totally different duties:

Is there a transparent winner between the 2? The reply isn’t any. Efficiency varies considerably by job:

• GPT-5 dominates in coding and reasoning
• GPT-4o holds its personal in content material creation and picture technology/evaluation
• Pace vs. Depth: GPT-4o delivers quicker responses, whereas GPT-5 generally hesitates between thorough evaluation and fast technology

Context issues: Keep in mind that GPT-4o is a yr older. Whereas GPT-5 advantages from newer coaching information and agentic optimizations, is it really groundbreaking in comparison with its predecessor? Not precisely.

Conclusion

Because the world calls for GPT-4o’s comeback, I wholeheartedly agree.

Whereas GPT-5 has improved since Day 1 (now outperforming its Day 3 outcomes), its rushed launch left customers struggling to adapt. The reality is, GPT-5 solely marginally surpasses GPT-4o on particular duties, making it painfully onerous to desert our beloved GPT-4o for one thing that feels merely “a tad higher.” Maybe OpenAI wanted extra rigorous testing earlier than launch. However now that it’s reside, we will solely watch its evolution.

As we speak? I’d signal any petition to convey again GPT-4o. ChatGPT has modified, and never for the higher. Let me know your ideas within the remark part.

PS: I took GPT 4o outputs from our earlier blogs: