Reasoning Levels

Some models spend extra time working through a problem before they answer. When the selected model supports it, 3sparks Chat lets you choose how much of that thinking to ask for.

Reasoning controls only appear when the connection and model actually support them. If you do not see the option, the selected model does not advertise reasoning.

Where to set it

• Chat settings — open the gear icon in a chat.
• LM Studio model pickers — the in-chat picker and LM Studio model management both expose a reasoning menu for models that support it.

Changes save immediately and apply to the next message you send. You do not need to reopen the chat.

LM Studio v1 models

For an LM Studio v1 connection, the levels offered are the ones the selected model advertises, so the list varies by model. Options come from Off, Low, Medium, High, and On, plus Model Default to leave the choice to the model.

Models that expose a plain on/off switch show On rather than graded levels. When the model publishes a description of its reasoning behavior, it appears beneath the picker.

OpenAI OSS connections

A connection using the OpenAI OSS API profile offers a fixed list: Provider Default, None, Minimal, Low, Medium, High, and Extra High. The choice is sent with the request as the reasoning effort.

Choosing a level

Higher levels mean more thorough thinking, slower replies, and more tokens spent. Lower levels return faster.

• Use higher levels for multi-step logic, debugging, planning, and analysis.
• Use lower levels or none for short questions, rewriting, translation, and casual chat.
• Start in the middle and adjust when an answer is thinner or slower than you want.

Reasoning output, when a model returns it separately from the answer, appears in its own collapsible section in the reply.

Switching models

Reasoning is a property of the model, not the chat. If you switch to a model that does not support the level you had chosen, 3sparks Chat clears the unsupported level instead of sending something the model would reject.

Troubleshooting

• If the reasoning control is missing, the selected connection or model does not support it. Confirm the model in chat settings.
• If a reasoning option errors, choose one the selected model actually lists.
• If replies are too slow, lower the level or pick a smaller model.
• If a level seems to be ignored, reopen chat settings to confirm it is still selected after a model switch.