News for hackers. Stuff that matters.

A comment on Qwen3-Max-Thinking. Parent comment

Re:Qwen3-Max-Thinking by storystarling on Monday January 26, 2026 @03:24PM (#46770987)
The closed nature is one thing, but the opaque billing on reasoning tokens is the real dealbreaker for integration. If you are bootstrapping a service, I don't see how you can model your margins when the API decides arbitrarily how long to think and bill for a prompt. It makes unit economics impossible to predict.
[ Parent | on HN ]
Re:Qwen3-Max-Thinking by TobTobXX on Monday January 26, 2026 @04:17PM (#46771672)
Doesn't ClosedAI do the same? Thinking models bill tokens, but the thinking steps are encrypted.
[ Parent | on HN ]
Re:Qwen3-Max-Thinking by Rastonbury on Monday January 26, 2026 @09:01PM (#46774555)
Destroying unit economics is a bit dramatic... you can chose thinking effort for modern models/APIs and add guidance to the system prompts
[ Parent | on HN ]
Re:Qwen3-Max-Thinking by zozbot234 on Monday January 26, 2026 @03:56PM (#46771402)
You just have to plan for the worst case.
[ Parent | on HN ]
Re:Qwen3-Max-Thinking by czl on Monday January 26, 2026 @09:08PM (#46774603)
FYI: Newer LLM hosting APIs offer control over amount of "thinking" (as well as length of reply) -- some by token count others by an enum (high low, medium, etc.).
[ Parent | on HN ]