The only one that is really close to Claude in performance is GLM-5.1. The others (Mimo, deepseek, etc..) looks good on paper but usually fails on a multi-step

The only one that is really close to Claude in performance is GLM-5.1. The others (Mimo, deepseek, etc..) looks good on paper but usually fails on a multi-step agentic orchestration. This is at least my experience with Claude Code as harness. Also, GLM pricing is not that far off from Claude. It's cheaper but not DeepSeek cheap.