GPT-5.2 reportedly makes basic spelling mistakes GPT-3.5 never did
OpenAI's 'smartest' model is regressing on basic spelling, users report.
Deep Dive
Viral user reports claim OpenAI's latest GPT-5.2 model is making frequent, basic spelling mistakes that earlier models like GPT-3.5 did not. This perceived regression in polish is sparking debate about whether it's a training flaw, a side effect of enhanced 'longer thinking' capabilities, or a decoding issue. The community is questioning if pushing for advanced reasoning has come at the cost of fundamental language accuracy.
Why It Matters
If true, it suggests a major trade-off in AI development where advanced capabilities may degrade basic, reliable performance.