Is it competitive overall or just competitive when compared to "U.S. Open-Weight models that are otherwise pretty decent at agentic coding"? The latter does not imply the former.