A codex with every kind of model Grok hot take: Super fast, average intelligence, but really fast GPT Sol: Smart, but talks too much nonsense, not very human-like Deepseek pro: Average intelligence, but great for the dirty work GLM 5.3: Smart, but loses IQ in long conversations (needs frequent session resets) Kimi 3: Smart but slow, not very practical Qwen 3.8 Max: Average intelligence, stingy, barely usable Minimax: Haven’t used it, no comment