I benchmarked Opus 4.8 vs. GPT 5.5 on 2 open source repos (www.stet.sh)

In a recent benchmark, Opus 4.8 has demonstrated superior performance compared to its predecessors and competitors, including GPT-5.5, by evaluating its capabilities on complex backend tasks from two open-source repositories (graphql-go-tools and sqlparser-rs) in Go and Rust. The analysis involved 50 real pull requests, revealing that Opus 4.8 not only leads on quality but also operates at a lower cost compared to GPT-5.5 and Opus 4.7. Specifically, while Composer 2.5 is significantly cheaper, it falls short in terms of code quality. The benchmark highlights that Opus 4.8 excels in creating maintainable and idiomatic code, achieving better "craft" scores across both languages, indicating improved clarity and simplicity. This progression is significant for the AI/ML community as it underscores the continuous advancements in AI models for software engineering tasks. Opus 4.8 solidifies its position as a refined tool that balances efficiency with quality, crucial for developers navigating complex coding challenges. This benchmark also calls attention to the importance of real-world testing over traditional binary assessments, suggesting that hands-on experience and local benchmarks are essential for evaluating AI capabilities in practical applications. Overall, the findings indicate a shift in preference towards Opus 4.8 for developers seeking reliability and cost-effective solutions in code generation tasks.