BullshitBench v2

Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better.

Claude Opus 4.8 by Anthropic holds the best BullshitBench v2 result among tracked models, at 95% (released May 28 2026). Higher scores are better on this benchmark.

1. Claude Opus 4.8 (Anthropic) — 95%; 1. Qwen3.8-Max (Qwen) — 95%; 3. Claude Sonnet 4.6 (Anthropic) — 91%; 4. Claude Opus 4.5 (Anthropic) — 90%; 5. Claude Opus 4.6 (Anthropic) — 87%.

84 tracked models have a published BullshitBench v2 score. Scores are gathered from BullshitBench, which runs the benchmark and publishes the full field, and are kept in step with its board.

Qwen3.8-27B by Qwen is the highest-ranked model with downloadable weights on BullshitBench v2, scoring 79% at rank 8 overall.