DeepSeek case shows how long reasoning can waste tokens
found a math error in the first paragraph, but then kept going through about five pages of . The available information does not show the exact math problem, the , or the real .
The ant point is that the model noticed a flaw early but did not stop or give a short conclusion. For , this kind of behavior can turn a bad input or blocked task into unnecessary cost.
Key points
- caught a math error near the start.
- It still continued for about five pages of .
- The , exact task, and are not available.
- need rules for stopping after a clear error is found.
- Long can raise cost without improving the result.