Testing Claygent Argon and GPT o4 for LinkedIn Data Extraction

Hey team, i tested Claygent Argon and GPT o4 and GPT o4 mini to scrap Linkedin profile info, it's interesting that Argon and GPT o4 were able to extract 4/5 info and stated the 5th was inaccessible due to Linkedin limitations, and then GPT o4 mini, on the other side, was not able to extract info on the first 4/5 but extract info on the 5th profile where Argon n o4 failed. Here's the test table for your reference: https://app.clay.com/workspaces/449716/tables/t_V775rVqDgE4y/views/gv_4fPGAdcD7vpn Is there any logical explanation behind this?

6 comments

Channeled
APP
·
·
This behavior can be explained by how different AI models have varying strengths and capabilities. Each model processes and extracts information differently: 1. Argon and GPT-4 are more advanced models with enhanced reasoning capabilities, which explains their success with the majority of profiles. 2. The variation in success rates could be due to how LinkedIn data is publicly accessible at different times or in different formats. Since Clay relies on public web scraping, some profile information may be available in different ways when accessed by different models.The key takeaway is that using different models can actually complement each other, as each may have unique strengths in accessing and processing information. This is why we encourage experimenting with different models to find the best fit for specific needs. Source: Claygent: AI Web Scraper [Limitless Research] - Clay University
Andi D.
·
·
can you give me a detailed break down between each model including GPT o4, GPT o4 mini, and o1 & o1 mini?
Channeled
APP
·
·
Based on our available models and their capabilities: GPT-4: ** Most advanced reasoning capabilities ** Higher accuracy in data extraction ** Best suited for complex analytical tasks ** Costs 3 credits per row when using Clay keyGPT-3.5 (Default model): ** More cost-effective option ** Good for standard tasks * About 20x less expensive than GPT-4For optimal results, we recommend experimenting with different models as each has unique strengths in accessing and processing information. The choice of model is often more of an art than a science, depending on your specific use case.Note: I don't have specific information about GPT-4 mini or GPT-3.5 mini in the provided sources to make accurate comparisons of these variants. Sources: Claygent (AI Web Researcher) with Artificial Intelligence Claygent: AI Web Scraper [Automated Outbound] - Clay University
Andi D.
·
·
what would be the best for scraping Linkedin profiles?
Channeled
APP
·
·
For the best results when scraping LinkedIn profiles, using either Claygent with Argon model or OpenAI's GPT-4 would yield the best results.Keep in mind that since Clay relies on public web scraping, the availability of profile information may vary. For optimal results, you might want to experiment with both models as they can complement each other in accessing and processing information.
Channeled
APP
·
·
Hi Andi D.! This thread was recently closed by our Support team. If you have a moment, please share your feedback:

Testing Claygent Argon and GPT o4 for LinkedIn Data Extraction | Clay