Table of Contents

Property ReasoningLevel

Namespace
LMKit.Agents
Assembly
LM-Kit.NET.dll

ReasoningLevel

Gets or sets the model-internal reasoning level for this execution.

public ReasoningLevel? ReasoningLevel { get; set; }

Property Value

ReasoningLevel?

Examples

Suppressing reasoning for a fast direct answer:

using LMKit.Model;
using LMKit.Agents;
using LMKit.TextGeneration.Chat;

using var model = LM.LoadFromModelID("qwen3:8b");
var agent = Agent.CreateBuilder(model).WithPersona("Classifier").Build();


// Force model-internal <think> emission OFF for this call only.
var result = agent.Run("Is this email spam? "Win a free iPhone now!"",
new AgentExecutionOptions
{
ReasoningLevel = ReasoningLevel.None
});

Remarks

Controls whether reasoning-capable models (Qwen3.5, GLM-4.7, Mistral Magistral, GPT-OSS, etc.) emit a <think>...</think> chain-of-thought block before their answer.

When null (default), the underlying conversation's ReasoningLevel flows through unchanged - identical default to driving MultiTurnConversation directly. PlanningStrategy and ReasoningLevel are orthogonal: planning controls prompt-engineered output structure, reasoning level controls model-internal <think> generation effort, and both compose.

Set explicitly to None when you want a fast direct answer without paying for chain-of-thought tokens. The agent's Content is already filtered against the runtime's segment classification, so leaving reasoning enabled does not leak chain-of-thought into the user-visible content - the only cost of leaving it on is generation time.

Share