I had previously noticed that Claude is significantly weaker than both Chatgpt and Gemini at search. Previously I was asking Claude, ChatGPT, and Gemini about a product listing on Amazon, and Claude was not even able to retrieve the item's Amazon listing page, whereas both chatGPT and Gemini were able to retrieve the Amazon page at least.
This time I asked Claude Opus 5.5 xHigh, ChatGPT (free), and Gemini 4 Argon High the exact same query about a character in a web comic. The character was blind.
Both Claude and ChatGPT were able to confirm that the character exists in the comic but were unable to find any sources to confirm the fact that the character was blind. This is despite multiple follow up prompts asking them to confirm that he was blind.
On the other hand, Gemini 4 Argon was able to find and confirm the fact that the character was blind on the first prompt, and also gave additional correct information about the character, indicating that it had found a correct factual source. No followup prompts were needed.
No comments:
Post a Comment