Posts tagged “benchmarks”
Opus felt like the worst model I tried for terminal work. It wasn’t.
Three coding agents doing server management from one infrastructure repo, and about 6,100 shell commands from their own transcripts. Opus 5.5 failed the least, on Windows and on my servers. It still felt like the worst, because Claude Code showed me every failure and loose end while Codex and ZCode hid most of theirs.