That poster isn't comparing models, he's comparing Claude Code to Cline (two agentic coding tools), both using Claude Sonnet 4. I was pretty much in the same boat all year as well; using Cline heavily at work ($1k+/month token spend) and I was sold on it over Claude Code, although I've just recently made the switch, as Claude Code has a VSCode extension now. Whichever agentic tooling you use (Cline, CC, Cursor, Aider, etc.) is still a matter of debate, but the underlying model (Sonnet/Opus) seems to be unanimously agreed on as being in a league of its own, and has been since 3.5 released last year.
