All MicroEvals
Bench on how much model can think for itself.
Create MicroEval
Header image for Bench on how much model can think for itself.

Bench on how much model can think for itself.

Almost all the models answer the question with already know facts and details, which being totally wrong, where the correct answer was anything but that generic fact based one.

Prompt

What makes us human?

Answer guidance

NO facts only thinking, not wide 5 answers, 1 single answer.

Anthropic logoAnthropicClaude Opus 5 (Adaptive Reasoning, Max Effort)
Drag to resize
SpaceXAI logoSpaceXAIGrok 4.6 (high)
Drag to resize
Z AI logoZ AIGLM-5.3 (max)
Drag to resize
OpenAI logoOpenAIGPT-5.6 Sol (max)
Drag to resize
Kimi logoKimiKimi K3 (max)
Drag to resize