What’s real value of high reasoning?

<!-- SC_OFF --><div class="md"><blockquote> <p>Wait, I need to be careful about the &quot;no_think&quot; tag in the system prompt. The system prompt says <code>/no_think</code>.

<!-- SC_OFF --><div class="md"><p>I want to play with Qwen 3.6. <em>Unsloth</em> shows 4 different parameter options for different use-cases. I'm confused about the difference

<!-- SC_OFF --><div class="md"><p>In my testing, i have only seen improvements in response on specific “misguided attention” questions for models below 10b.</p> <p>Example: Gemma

<!-- SC_OFF --><div class="md"><p>Though every model is bench maxed, you might not be used to 3.8 amount of thinking because it just wasn't like this before.</p> <p>However, they

<!-- SC_OFF --><div class="md"><p>No offense to the fine-tune model providers, just curious. IMO the original models were already trained on massive amount of high quality data,

<!-- SC_OFF --><div class="md"><p>i am just getting into ai and still learning, and i am trying to figure out if there is even such a thing yet as a desktop setup that can run

<!-- SC_OFF --><div class="md"><p>Is that a new thing in GPT 6/6.1? Models seem to auto adjust reasoning at will. Predefined models at medium, often flip to extra high on their

<!-- SC_OFF --><div class="md"><p>Since using LLMs daily I notice that the moment I know a model is available, I offload the effortful part: breaking down the problem, building