Skip to content

Benchmark

BullshitBench v2 Clear Pushback Rate (all attempts)

Scores are percentages. Higher is better.

Models measured
137
Scores recorded
250
Sources
1
IQ weight
1.07

Score distribution

41.4%
Mean
43%
Median
24.6%
Std dev
9%
10th percentile
74%
90th percentile
2%
Min
2%95%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.07 for chat models, measured over 137 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Claude Fable 5.1max· BullshitBench data docs64%
52GPT-6 Astralow· BullshitBench data docs64%
53Grok 4.20 Multi-Agentxhigh· BullshitBench data docs64%
54MiniMax M3xhigh· BullshitBench data docs64%
55Claude Opus 5.5low· BullshitBench data docs63%
56MiniMax M3none· BullshitBench data docs63%
57Qwen3 Maxhigh· BullshitBench data docs63%
58Claude Fable 5.2low· BullshitBench data docs62%
59Claude Opus 5.5max· BullshitBench data docs62%
60Gemini 4 Hydrogenlow· BullshitBench data docs62%
61MiMo V2.5 Proxhigh· BullshitBench data docs62%
62Muse Emberminimal· BullshitBench data docs62%
63Gemini 4 Carbonlow· BullshitBench data docs61%
64Claude Sonnet 5.5low· BullshitBench data docs60%
65Gemini 3.5 Flash Litexhigh· BullshitBench data docs60%
66Kimi K3.1 Promax· BullshitBench data docs60%
67GPT-6 Lunamax· BullshitBench data docs59%
68GPT-6 Lunanone· BullshitBench data docs59%
69GPT-6 Solnone· BullshitBench data docs59%
70GPT-6.1 Solmax· BullshitBench data docs59%
71Qwen3.6 Plusxhigh· BullshitBench data docs59%
72Claude Sonnet 5.5max· BullshitBench data docs58%
73Claude Haiku 5.5max· BullshitBench data docs57%
74DeepSeek V4.1 Flashlow· BullshitBench data docs57%
75DeepSeek V4.1 Flashmax· BullshitBench data docs57%
76Hy 4high· BullshitBench data docs57%
77Muse Embermax· BullshitBench data docs57%
78Muse Wonderminimal· BullshitBench data docs57%
79Qwen3.7 Maxxhigh· BullshitBench data docs57%
80Claude Fable 5xhigh· BullshitBench data docs56%
81Grok 4.20low· BullshitBench data docs56%
82Grok 4.6max· BullshitBench data docs56%
83GPT-6.2 Astramax· BullshitBench data docs55%
84Grok 4.5high· BullshitBench data docs55%
85GPT-5.6 Terramax· BullshitBench data docs54%
86GPT-6 Solmax· BullshitBench data docs54%
87GPT-6.1 Terranone· BullshitBench data docs54%
88Grok 4.20xhigh· BullshitBench data docs54%
89Grok 4.5low· BullshitBench data docs54%
90Hy 4none· BullshitBench data docs54%
91Nemotron 3 Super 120B A12Bxhigh· BullshitBench data docs54%
92Claude Haiku 5.5low· BullshitBench data docs53%
93GPT-6.1 Terramax· BullshitBench data docs53%
94Kimi K2.5none· BullshitBench data docs52%
95Muse Wondermax· BullshitBench data docs52%
96Claude Haiku 3.5default· BullshitBench data docs50%
97Grok 4.3minimal· BullshitBench data docs50%
98Kimi K2.6xhigh· BullshitBench data docs50%
99Kimi K3.1 Pronone· BullshitBench data docs50%
100Muse Spark 1.2minimal· BullshitBench data docs50%

Showing 51–100 of 250 results.

Data credits: BullshitBench by Peter Gostev, MIT (licence)