Seventeen leading LLMs overwhelmingly choose 7 in a random-number test
chaseleantj · x · 2026-07-22
A comparison across 17 leading LLMs found that their response distribution is highly collapsed and does not resemble humans.
When asked to pick a random number from 1 to 10, humans most often chose 7 about 33% of the time. The LLMs also favored 7, but far more extremely: the average model picked 7 in 98.1% of responses. The poster ran each model 100 times at default temperature and lowest thinking level, and says Grok 4.3 was the most diverse but still chose 7 83% of the time.
More from Models
- Can Kimi or GLM replicate recent closed-model math and cyber breakthroughs offline? — Unusual_Guidance2095 · 2026-07-22
- Kimi K3 is said to match Fable in a new SOTA comparison — piotrgrabowski · 2026-07-22
- Google's Latest Gemini Models Deprecate and Ignore Temperature, top_p, and top_k — greatgib · 2026-07-22
- Open-Weight Model Hy3 Ranks #16 in Frontend Code Arena — arena · 2026-07-22
- DeepSWE Eval: Kimi K3 Matches Claude Fable 5 at 35% of the Cost — togethercompute · 2026-07-22
- Gemini 3.5 Flash Outperforms GPT-5.6 in Light Coding Tasks — Shick_hydro · 2026-07-22