Assume a task normally requires 8000 thought tokens to complete if no budget is set. Now, if the budget is set to 3000 tokens, will the task fail due to insufficient thought tokens, or will it still be completed, but with lower performance compared to the 8000-token version?
@hong_jackey, the model will still complete the task within the thinking budget, but as the complexity of the task at hand increases, the performance might benefit from increasing the thinking budget.