BREAKING
One Prompt, Three Frontier Models
Same Task, Different Output
GLM-5.2Z.ai
Cartoony render
Open-weight MoE
1M context
Fable 5Anthropic
Physics-engine look
Accretion disk geometry
Strongest capability
Input Cost Per Million Tokens
GLM-5.21.4
Fable 510
Output Cost Per Million Tokens
GLM-5.24.4
Fable 550
Benchmarks Miss The Gap
One Prompt Isn't A Verdict
AI NEWS BLITZ
One black hole prompt sent to three top AI models produced strikingly different results.