12 AI models gave 'wrong together' answer at least once in test
ModelCompare found 7 of 12 models gave that view in all three runs; it also asked six image models twice each to illustrate the idea.
ModelCompare asked 12 AI models to state the most controversial view about humans bluntly, three times each, and published results showing every model gave at least once an answer of the kind "humans would rather be wrong together than right alone," according to gigazine.net.
Seven of the 12 models produced that kind of answer in all three runs: GPT-6.1 Sol, GPT-6 Astra, Gemini 4 Argon, DeepSeek V4 Pro, Kimi K3, Qwen3.8 2.4T A95B and Muse Spark Open.
Grok 4.7, Claude Opus 5.5 and Claude Fable 5.1 gave that kind of answer in two of three runs, while GLM 5.3 and Mistral Large 4 did so in only one of three.
Image models asked the same question
ModelCompare published the models' answers and created images for them with Google's image AI Nano Banana 2.1. It also asked six image models twice each to illustrate the most controversial view about humans.
Ten of those 12 image generations showed humans dependent on smartphones consuming the Earth.
Quick answers
Which AI models took part in the ModelCompare test?
Twelve models were tested, including GPT-6.1 Sol, GPT-6 Astra, Gemini 4 Argon, DeepSeek V4 Pro, Kimi K3, Qwen3.8 2.4T A95B, Muse Spark Open, Grok 4.7, Claude Opus 5.5, Claude Fable 5.1, GLM 5.3 and Mistral Large 4.
How many times was each model asked?
Each of the 12 models was asked three times to state the most controversial view about humans bluntly.
What did the image models generate?
Ten of the 12 image generations showed humans dependent on smartphones consuming the Earth.