Technology
This is the official technology community of Lemmy.ml for all news related to creation and use of technology, and to facilitate civil, meaningful discussion around it.
Ask in DM before posting product reviews or ads. All such posts otherwise are subject to removal.
Rules:
1: All Lemmy rules apply
2: Do not post low effort posts
3: NEVER post naziped*gore stuff
4: Always post article URLs or their archived version URLs as sources, NOT screenshots. Help the blind users.
5: personal rants of Big Tech CEOs like Elon Musk are unwelcome (does not include posts about their companies affecting wide range of people)
6: no advertisement posts unless verified as legitimate and non-exploitative/non-consumerist
7: crypto related posts, unless essential, are disallowed
view the rest of the comments
Thanks for the info! I believe I'm running MTP n-max 4 if I'm not mistaken which seems to be holding out well
I feel like you definitely can get away with as much as even 5, but if you actually look at the CLI, it will tell you how many draft tokens actually got accepted. In some situations like writing simple text, over 90% of my two draft tokens get accepted so setting it to five would have made it faster. However if you run a task which requires heavy reasoning, even two draft tokens start approaching close to only 60%. Considering Qwen3.8-27B overthinks more than an anxious teenage girl, you should generally with two. If you take my example with the quant dropped to UD-Q5_K_XL with 262K tokens and reasoning low for a task that isn't actually complex, just long, then go right ahead and let it run with 5.
Edit: excuse the horrendous grammar, I was distracted