SlopCodeBench October 2026: GPT-6 Astra leads strict pass rate, XHigh effort yields diminishing returns

·2026.10.16 04:09

Key point

GPT-6 Astra achieved the highest strict pass rate of 16.3% on SlopCodeBench, while higher effort settings like XHigh often increased costs without improving or even degrading strict pass rates.

1 / 6

Details

GPT-6 Astra XHigh recorded the highest strict pass rate at 16.33% (32/196 checkpoints), narrowly beating GPT-6 SOL Medium at 15.31%.
Higher effort settings (XHigh) did not reliably improve strict pass rates; Opus 5.5 and GPT-6 SOL performed worse on XHigh than Medium, with XHigh costing 2.11x to 4.22x more.
Final code quality remains poor, with 1,435 to 2,831 failed test cases in the final checkpoint across models, indicating significant regression issues.
Opus 5.5 XHigh produced the cleanest code with the lowest slop coverage (77.75%), despite having fewer strict passes than its Medium counterpart.

benchmark ai-coding model-evaluation slopcodebench gpt-6 opus-5.5

Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.