xhigh
A high tier in an inference-effort grading system, typically positioned above high and near the maximum but not the absolute upper bound, representing a relatively large computational budget the model may spend before returning an answer.
Timeline (last 90 days)
In most reasoning-intensity tiering systems, the levels typically progress from low, medium, high to xhigh, with xhigh being a near-top tier rather than an absolute ceiling.
The SlopCodeBench October test report shows that after adjusting reasoning effort from medium to xhigh, cost rose to 2 to 4 times, while the number of checkpoints passed did not improve reliably.
On checkpoint pass rate, xhigh showed no statistically reliable improvement; on some tasks the two were on par, and on individual tasks xhigh was even slightly lower.
Checkpoint evaluation inherently favors breadth coverage: quickly identifying and satisfying as many checkpoints as possible is more worthwhile than digging deep into a single hard point; under the xhigh level, the model invests too much reasoning budget in a single sub-problem, which may leave the remaining checkpoints untouched and lose points.
The report points out that the extra cost brought by xhigh does not correspond to a reliable gain in pass rate, constituting diminishing marginal benefit and even being negative in some scenarios.
The title of the SlopCodeBench 2026 leaderboard includes "xhigh considered harmful", meaning the highest reasoning effort level (xhigh) lowers the final score.
When offering reasoning effort options, the highest level (xhigh) should not be recommended by default; developers should choose the appropriate level based on the nature of the task.
All Facts (7)
In most reasoning-intensity tiering systems, the levels typically progress from low, medium, high to xhigh, with xhigh being a near-top tier rather than an absolute ceiling.
50%UnverifiedThe SlopCodeBench October test report shows that after adjusting reasoning effort from medium to xhigh, cost rose to 2 to 4 times, while the number of checkpoints passed did not improve reliably.
50%UnverifiedOn checkpoint pass rate, xhigh showed no statistically reliable improvement; on some tasks the two were on par, and on individual tasks xhigh was even slightly lower.
50%UnverifiedCheckpoint evaluation inherently favors breadth coverage: quickly identifying and satisfying as many checkpoints as possible is more worthwhile than digging deep into a single hard point; under the xhigh level, the model invests too much reasoning budget in a single sub-problem, which may leave the remaining checkpoints untouched and lose points.
50%UnverifiedThe report points out that the extra cost brought by xhigh does not correspond to a reliable gain in pass rate, constituting diminishing marginal benefit and even being negative in some scenarios.
50%UnverifiedThe title of the SlopCodeBench 2026 leaderboard includes "xhigh considered harmful", meaning the highest reasoning effort level (xhigh) lowers the final score.
50%UnverifiedWhen offering reasoning effort options, the highest level (xhigh) should not be recommended by default; developers should choose the appropriate level based on the nature of the task.
50%