3 ms·
While GPT-5.2 scores well on benchmarks, human preference is important for OpenAI’s consumer focused products.
by reed1234 10mo ago
While GPT-5.2 scores well on benchmarks, human preference is important for OpenAI’s consumer focused products.
- aeonfox 10mo agoArena Overview section is heavily biased towards languages. grok-4.1-thinking is worse than claude-opus-4-5-20251101-thinking-32k on every non-language metric by a large margin but somehow ranks higher overall, maybe because opus is way worse Spanish and Korean?