Reasoning Budget
A core hyperparameter in large reasoning models that controls the depth of internal deliberation. It sets an upper limit on the number of internal reasoning tokens the model may generate before producing a final answer, and is typically configured in tiers such as Low, Medium, High, Extra High, and Max.
Timeline (last 90 days)
The Reasoning Budget is jointly defined by three dimensions—token count × cost × maximum latency—and exposes a "budget exceeded" flag.
The engineering significance of the reasoning budget lies in: latency and cost often trade off in opposite directions (large models are more expensive but may get it right in one shot, while small models are cheap but may retry repeatedly and end up slower), and looking at either metric alone will lead to wrong decisions.
The three dimensions of depth, width, and feedback share the same reasoning budget pool; increasing depth means reducing opportunities for width exploration, and investing in feedback iterations may sacrifice the depth of a single reasoning pass.
All Facts (4)
The Reasoning Budget is jointly defined by three dimensions—token count × cost × maximum latency—and exposes a "budget exceeded" flag.
50%UnverifiedThe engineering significance of the reasoning budget lies in: latency and cost often trade off in opposite directions (large models are more expensive but may get it right in one shot, while small models are cheap but may retry repeatedly and end up slower), and looking at either metric alone will lead to wrong decisions.
50%UnverifiedThe three dimensions of depth, width, and feedback share the same reasoning budget pool; increasing depth means reducing opportunities for width exploration, and investing in feedback iterations may sacrifice the depth of a single reasoning pass.
50%Unverified现代推理模型通常具备推理预算机制，为每次推理分配一定量的令牌配额
50%