Claude 3.5 sonnet Vs GPT-4o: Key details and comparison
In this article, you will get to learn about the most commonly used AI model, a full comparison of Claude 3.5 sonnet vs GPT-4o along with their capabilities and use cases.
Claude 3.5 and GPT-4o are the most commonly used AI models right now.
Whether for coding, writing, or researching, both models are on par with their capabilities.
However, the internet is divided when it comes to choosing the best one.
So, we will use those opinions, community-run data, personal experiments, and standard benchmarks to compare the two and provide you with a rundown of their capabilities, use cases, and performance comparison.
TLDR;
Claude 3.5 and GPT-4o are both great models that can help with our day-to-day tasks. While Claude is better at logic, GPT-4o provides better results for complex reasoning.
When comparing Claude AI with ChatGPT, if you are looking for an AI that can help you be more creative with less generic responses, then Claude is a better fit. But for coding, mathematical problems, and analysis, ChatGPT should be your choice.
What is Claude?
Claude is both an AI assistant and the name of the large language model that powers it, developed by the Anthropic team. Claude is said to be an ethical alternative to ChatGPT and has been built by ex-OpenAI members.
You can read more about their “Constitutional AI” approach here. With the help of Claude, you can write prompts in natural language and get help with summarization, decision-making, code writing, research, Q&A, editing, and more.
Claude’s current models
Claude has a total of 5 models in the Claude 3 model family. These are known for near-instant results, strong vision capabilities, improved accuracy, long context, and near-perfect recall.
Here are all the models present in the Claude 3 model family, along with their APIs:
These are so far the fastest, most intelligent, and best models by Claude. Here’s a detailed comparison from the Anthropic team on their models:
All about Claude 3.5 models
In the 3.5 models series of Claude, there are two models:
Claude 3.5 Sonnet – The most intelligent model
Claude 3.5 Haiku –The fastest model
Claude 3.5 Sonnet is the most widely used model, and when we did a comparison of Claude 3.5 Sonnet vs Opus, we found that it is more intelligent, faster, and also more cost-effective.
According to an internal agentic coding evaluation by the Anthropic team, Claude 3.5 Sonnet solved 64% of problems, outperforming Claude 3 Opus which solved 38%.
To use Claude 3.5 sonnet, go to claude.ai/new, simply sign up and you can start using it.
Now that we know Claude 3.5 Haiku and Sonnet are Claude’s fastest and most intelligent models, this brings up the next question, which is, “Is Claude 3.5 free?”
Claude 3.5 Sonnet is free to use, but it comes with some limitations:
You may hit usage limits with around ten prompts.
You can use Claude 3.5 Sonnet model as an API, for $3 per million tokens input and $15 per million tokens output.
Whereas, for Claude 3.5 Haiku, you need to subscribe to the Claude Pro plan, which costs $20 per month.
The release of Claude 3.5 sonnet also saw the introduction of Artifacts.
When you ask Claude to generate content like code snippets, or text documents, these Artifacts appear in a dedicated window alongside the conversation as shown in the image below.
This creates a dynamic workspace where you can see, edit, and build upon Claude’s creations in real-time.
What is ChatGPT?
ChatGPT is an AI chatbot developed by the OpenAI team.
It can help in generating code, text, images, and videos with prompts. (If you are a developer and want to write code with the help of ChatGPT, here’s a helpful article).
The models that power ChatGPT, are developed using three primary sources of information: (1) information that is publicly available on the internet, (2) information that OpenAI partners with third parties to access, and (3) information that OpenAI users or human trainers and researchers provide or generate.
All available OpenAI models
OpenAI models and its APIs can be used for a variety of tasks, here’s a detailed table on the available models and their descriptions:
What is GPT-4o?
GPT-4o is OpenAI’s high intelligence model and was available for preview on May 13th, 2024. The 'o' in GPT-4o stands for "omni."
It accepts any combination of text, audio, image, and video as input and generates any combination of text, audio, and image outputs.
It can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in a conversation.
How GPT-4o works exactly is not known, but according to OpenAI's announcement video, it is a single neural network that was trained on text, vision, and audio input.
When it comes to pricing, GPT-4o can be used in ChatGPT with a monthly subscription of 20$ and $2.50 / 1M input tokens for the API.
Comparing GPT-4o with other OpenAI models
The latest model that OpenAI has to offer is o1, which is designed to think, help in complex tasks, and solve harder problems than previous models in science, coding, and math.
When we do a quick comparison of OpenAI o1 vs 4o, o1 doesn't yet have many of the features that make ChatGPT useful, like browsing the web for information and uploading files and images, but can help with reasoning tasks.
For daily and more common uses, GPT-4o turns out to be more capable.
The next thing that you must have thought is,“Is GPT-4o better than GPT-4?” . While some users on the OpenAI forum say that GPT-4 is better, when it comes to benchmarking performance, contextual understanding, speed, cost, and language support, GPT-4o is ahead and provides much better responses.
Claude 3.5 Sonnet vs GPT-4o
We will be comparing Claude 3.5 sonnet vs GPT-4o in terms of the following benchmarks:
Graduate Level Reasoning
Math problem solving
Latency & Speed
Accuracy
Usage
Pricing
GPQA, Diamond
The GPQA, Diamond is a benchmark designed to evaluate an AI’s graduate-level reasoning capabilities. In this, Claude 3.5 Sonnet leads with a 59.4% score on 0-shot CoT GPQA, while GPT-4o has a 53.6% score on zero-shot CoT.
MATH
MATH is a benchmark designed to evaluate AI’s ability to solve complex mathematical problems. In this, GPT-4o leads with a 76.6% score on 0-shot CoT, while Claude 3.5 Sonnet has 71.1% score on zero-shot CoT.
Latency & Speed
When it comes to latency and speed, GPT-4o is ahead of Claude 3.5 Sonnet according to co-founder from Keywords AI.
GPT-4o’s average latency is 24% faster than Claude-3.5-Sonnet (7.5226s vs 9.3055s)
GPT-4o even has more output tokens than Sonnet (431 tokens/request vs 260 tokens/request)
GPT-4o’s average time to first token(TTFT) is 2x faster than Claude-3.5-Sonnet (0.5623s vs 1.2341s)
GPT-4o’s average speed is also 2x faster than Claude-3.5-Sonnet (56T/s vs 28/2T/s)
Accuracy
To see how accurate the models are, I did a quick test on Pieces to check the response to Can you help me write a concise article on Pwn Request for GitHub Actions, here are the results:
Claude 3.5 Sonnet
GPT-4o
In this, Claude 3.5 Sonnet gave an incorrect response, while GPT-4o correctly mentioned that it is a vulnerability. (You can read about Pwn Request in this article).
This shows that GPT-4o has higher accuracy than Claude 3.5 Sonnet.
Usage
While comparing Claude vs OpenAI in terms of usage, Claude is often considered better for reasoning tasks and has a more human-like interaction style, while OpenAI is more versatile, along with better complex reasoning.
Pricing
The final comparison is the ChatGPT vs Claude subscription. ChatGPT comes in three subscription tiers: free, plus at $20 per month with certain limits, and pro at $200 per month.
Claude comes in four subscription tiers: free, pro at $18 per month, team at $25 per person per month, and custom pricing for the enterprise.
The bottom line
Throughout this article, we came across many concepts like prompt engineering, artifacts, model performance, and analysis.
If you would like to learn more about these, here is a helpful set of resources:
This article was first published on September 16th, 2024, and was improved by Haimantika Mitra as of January 3rd, 2025, to improve your experience and share the latest information.
