SlopCodeBench 10/2026 – Opus 5.5, Astra wins overall, xhigh considered harmful
Contribute to humanlayer/advanced-context-engineering-for-coding-agents development by creating an account on GitHub.
AI Summary
The SlopCodeBench update for October 2026 highlights the performance of various coding models, with gpt-6 astra achieving the highest score by improving 1.5% over gpt-5.5. The benchmarks focus on strict and core pass rates, revealing that frontier models had strict pass rates between 12.2% and 16.3%, while core pass rates were significantly higher, at 65.8% to 74.0%. Notably, medium effort performed better than XHigh for some newer models, despite the latter being more expensive.


