My prime models are GPT-5.2 Codex XHigh, Opus 4.5 thinking, Gemini 3 pro, but also Kimi 2 thinking and now Kimi 2.5 for plan reviews or brainstorming. Kimi often find one brillant idea or dig one aspect that was overlooked or missed by the Big 3.
My secondary is GLM 4.7 and it's actually very good. GLM 4.6 was a mid-secondary, 4.7 is a top one.
My prime models are GPT-5.2 Codex XHigh, Opus 4.5 thinking, Gemini 3 pro, but also Kimi 2 thinking and now Kimi 2.5 for plan reviews or brainstorming. Kimi often find one brillant idea or dig one aspect that was overlooked or missed by the Big 3.
My secondary is GLM 4.7 and it's actually very good. GLM 4.6 was a mid-secondary, 4.7 is a top one.
I've heard really good things about Kimi, and now I see Qwen3-coder-next https://qwen.ai/blog?id=qwen3-coder-next is getting a lot of love.
So many choices, good problem to have.
Qwen Code with its free daily 2K requests is a good candidate to put inside a Ralph loop.
For now, let's try Opus 4.6 and Codex 5.3!