tech
December 28, 2025
Vibe Check: GPT-5.2 Is an Incremental Upgrade
OpenAI's latest model update excels at instruction-following and extended tasks, but don't expect it to surprise you
TL;DR
- GPT-5.2 is an incremental update to OpenAI's models, offering improvements over GPT-5.1.
- It excels in long-running, complex knowledge work and analytical tasks, demonstrated by its performance on profit and loss statement analysis.
- OpenAI claims GPT-5.2 scores 70.9% on GDPVal, a significant increase from GPT-5.1's 38.8%.
- In writing tests, GPT-5.2 scored 74%, matching Sonnet 4.5 but trailing Opus 4.5.
- The model is better at following instructions, like being sarcastic, compared to Claude Haiku.
- GPT-5.2 is less resourceful than Opus 4.5 in tasks requiring independent problem-solving, such as searching email for location data.
- While GPT-5.2 is beneficial for analytical tasks within ChatGPT Pro, users should temper expectations for drastic improvements in everyday chat.
- Opus 4.5 remains preferred for tasks demanding creativity, intelligence, and autonomy.
Continue reading
the original article