Burkov: closed LLM providers bill you for hidden thinking tokens you can never verify

burkov · x · 2026-09-19

Andriy Burkov raises a pointed concern: when using closed-weight LLMs, users pay for thinking tokens that providers never send back to the caller, so there's no way to verify they were actually generated. Advertised prices like $2/1M input and $10/1M output could be underpriced, with hidden reasoning tokens padding the bill. He also questions why some models force reasoning to stay on at least at a minimal effort level.

Related event: Burkov questions opaque thinking-token billing in closed LLMs(2 posts)→

Original post →

More from Models

Models channel →