Brief, mostly signal
Opus 5.5 was not always the cheapest model to finish coding jobs
A Reddit post compares Claude Code and Codex on four jobs of varying size. Only the title was readable, so the cost data is unverified.
What the research found
Reddit post claims Opus 5.5 was never the cheapest model to finish four coding jobs of varying size in Claude Code and Codex, but cached text is only the title; Reddit blocked full fetch, and no cost data is verifiable.
Claims and how well the evidence supports them
- Opus 5.5 was never the cheapest model that got the job done across four coding jobs of different sizes 7%
- The test ran jobs through both Claude Code and Codex 5%
Evidence
- reddit.com https://reddit.com/r/ClaudeAI/comments/1xbrcvi/i_ran_4_coding_jobs_of_different_sizes_through
Actionability rated low.