Been testing out GLM-4.7, and it's pretty impressive. I had it evaluate an nftables ruleset I put together with help from ChatGPT and Gemini.

My take:
ChatGTP 5.2 Thinking (Exntended thinking) >= GLM-4.7 (Deep Think) > Gemini 3.0 Pro

Started a new project with Rust + Elixir. Using Gemini 3 Pro for writing specs, and Claude Code for implementation. But the Claude Pro plan might be a bit tight.