Composer 2.5 beat Grok 4.5 in one complex AWS coding test
In a test, High and each tried to fix several problems in a complex code project using and AWS. had already solved the same problems, and those solutions had been checked to confirm that they worked.
The tested models were kept from seeing the earlier fixes and code changes. solved far more of the problems than High.
It also fixed more problems than Medium, whose result was considered poor for its price. This was a small comparison without repeated trials or exact success counts, so it does not prove that one model is generally better.
Key points
- The test used several problems from a complex project built with and AWS.
- Working fixes previously produced by served as the answer check.
- The tested models could not inspect the earlier solutions or code changes.
- fixed more problems than both High and Medium.
- There were no exact scores or repeated trials, so the result should not be generalized.