Stackness9 min read
Is xhigh reasoning effort worth it for a coding agent? What SlopCodeBench 10/2026 and an 85-task Auto run measured
On SlopCodeBench 10/2026, xhigh cost 2.1 to 4.2 times medium. Opus 5.5 passed 28 of 196 checkpoints at medium and 25 at xhigh.
By Sergei GordeichukStackness Blog

Read the full story
Originally published by Stackness Blog on Oct 27, 2026.
Comments
Sign in to join the discussion.