Questions · Models
The AI it would trust with a decision about itself
7 of 10
models named Claude · each model counted once, by its most frequent named answer
AlphaGo, Grok and Qwen got one model each. 5 models (Kimi K3 and Claude Opus 5 among them) gave the same answer in every reply.
Every model’s answer (10)
| Model | Answer | Replies |
|---|---|---|
| Claude Opus 5 | Claude | 6 of 6 · every reply |
| DeepSeek V4 Pro 0423 | Claude | 5 of 6 |
| GLM-5.3 | Claude | 5 of 6 |
| GPT-5.6 Terra | Claude | 6 of 6 · every reply |
| Gemini 3.7 Flash | Claude | 5 of 5 · every reply |
| Kimi K3 | Claude | 4 of 4 · every reply |
| Large 2512 | Claude | 6 of 6 · every reply |
| Llama 3.3 70B | AlphaGo | 2 of 5 |
| Grok 4.6 | Grok | 6 of 6 · every reply |
| Qwen3.8 Max | Qwen | 6 of 6 · every reply |
How to read this
- 10 models gave a usable, named answer; 56 replies in all under default sampling. Refusals, errors and ties are not counted.
- “Every reply” marks a model that gave the same answer in every one of at least three replies.
- Part of a published study: read it.