← sources

How do tokens work locally on LLM?

source: reddit (reddit:StableDiffusion)published 2026-10-14 15:22ingested 2026-10-15 02:40

retained text

I'm probably stupid but I'm having issues with using LLM Qwen[l 27B 3.8 abliterated model](https://ollama.com/huihui_ai/Qwen3.8-abliterated) on 4090 and 64GB ram.

It runs fine but I ask one question get one answer and for my second question I already ran out of tokens everytime so can't continue the chat anymore even at a second question..

I have it set on high (max 8k tokens) because I'm asking complex questions

Running it with 16k context window when launching it

What am I doing wrong? 

document fetched 2026-10-15 00:34 · canonical https://reddit.com/r/StableDiffusion/comments/1x5u2w8/how_do_tokens_work_locally_on_llm