this! though glm-5.2 feels more like an sonnet alternative to me (less coding & more talking and brainstorming)

deepseek-v4-pro/flash do feel quite versatile given the token prices and great caching

also kimi (k2.7-code) might be on par (if not better) with gpt5.5 in adversarial reviewing (codes, texts, etc). though it's super critical (as it like to adopt a critical realist perspective to everything)