OpenAI and Anthropic face backlash over distillation claims and hidden reasoning

max_paperclips · x · 2026-07-26

OpenAI and Anthropic are being criticized for a contradiction around distillation: some labs said their models were impossible to distill, while now they are said to be using distilled models themselves.

The thread also claims that the “reasoning” shown in the UI is not the raw chain of thought, but a filtered summary of an encrypted internal reasoning process.

Original post →

More from Models

Models channel →