🤖 Claude beat ChatGPT in a blind test of 6 tasks
The author of the channel “Misha, let's do it again” published a blind comparison of GPT-5.1 High and Claude Fable 5.1 on iXbt Live: each run was a new chat with no memory or personalization, with hidden traps. On 10-point ratings, Claude scored 54 points versus 47 for ChatGPT, which hallucinated in an incident report and missed a deadline.
🌍 The blind methodology without a “warmed-up” context shows: on office tasks, top models are on par, and the real differentiator is rare failures like data fabrication, not raw leaderboard scores.
👤 In the code, both models found all 5 planted bugs, but Claude's script for the author's measurement was faster — 0.14 seconds versus 0.37 seconds. In the tables, Claude had live Excel formulas, while ChatGPT had web search and an audit of 843 edits.
Source 1: https://www.ixbt.com/live/sw/chatgpt-protiv-claude-sravnivaem-topovye-neyronki-na-6-zadachah.html