Model techniques map
Techniquesinference & servingreasoning control

specific method · filed under inference & serving

Default thinking mode

The model operates in thinking mode by default, producing reasoning content before its response.

Also called Default thinking mode for generation, thinking mode by default, think mode, thinking mode, Thinking mode (default reasoning mode).

sources
8
models
6
labs adopt it
3
strongest
core

How sources treat it

One count per evidence span, weakest treatment to strongest.

evaluated 1used 1default 5core 1

Documented in

Evidence

8 spans quoted from the sources, strongest treatment first.

We introduce a thinking mode [OpenAI, 2024] to Gemma 4 models. By outputting a reasoning trace before the response, models demonstrate improved capabilities in reasoning-heavy domains such as mathematics and coding.

corepost trainingin Gemma 4Google DeepMind

Qwen3.5 models operate in thinking mode by default, generating thinking content signified by thinking... before producing the final responses.

defaultpost trainingin Qwen3.5-122B-A10BQwen

Qwen3.5 models operate in thinking mode by default, generating thinking content

defaultinference servingin Qwen3.5-397B-A17BQwen

Qwen3.5 models operate in thinking mode by default

defaultinference servingin Qwen3.5-35B-A3BQwen

Qwen3.8-Flash-Next models operate in thinking mode by default

defaultunclearin Qwen3.8-Flash-NextQwen

Qwen3.6 will think by default before response.

defaultinference servingin Qwen3.6-27BQwen

All models were evaluated in think mode

usedevaluation onlyin GLM-5.2Z.ai

The performance of non-thinking mode is slightly worse than the thinking mode, but still competitive.

evaluatedinference servingin DeepSeek-V3.2DeepSeek

Filed alongside

Other methods under inference & serving :: reasoning control.