silenceimpaired

• The best front end GUI

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 11 comments

• Qwen 3.5 27b dense vs 122b MoE

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 1 comments

• Qwen 3.5 27b Dense vs Qwen 3.5 120b MoE

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 6 comments

• Too many large MoEs, which do you prefer for general instruction following/creative endeavors? (And why)

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 36 comments

• Okay, I'm overthinking... yes, yes you are Qwen 3.5 27b.

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 15 comments

• Dear model creators…

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 6 comments

• Deepseek architecture, but without all the parameters

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 34 comments

• Why did LLM360's K2-V2 Instruct not get picked up by finetuners?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 8 comments

• What happens when you load two models and let each model take a turn generating a token?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 16 comments

• There has been a lot of efforts in the past to improve quantization due to the size of dense models… are we likely to see improvements like pruning and/or distillation with the uprise of huge MoEs?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 16 comments

• We know the rule of thumb… large quantized models outperform smaller less quantized models, but is there a level where that breaks down?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 85 comments

• K2-V2 - 70B and creative writing

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 14 comments

• Deepseek 700b Bitnet

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 19 comments

• Is there a local tool that lets me have the LLM process a large swath of text based on a prompt?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 3 comments

• A return to dense models?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 20 comments

• ChatGPT, Gemini, Grok, and OpenRouter… only cost you double the price of RAM and GPUs!

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 10 comments

• Fine tuning - is it worth it?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 22 comments

• If you could have one LLM distilled to a smaller size, which would model would you pick and what size(s) would you pick?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 35 comments

• 2-3 years out from now NPCs in Games won't be one dimensional

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 33 comments

• When AI defines our world, what will define us?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 9 comments

• Help me fish...

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 7 comments

• Tools and practices for a stable rolling release distro like arch

  Posted by silenceimpaired@reddit | linux | View on Reddit | 13 comments

• Anyone starting creative writing fine-tuning on Seed-OSS-36B-Base-woSyn? I've thought about it.

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 6 comments

• Is EXL3 doomed?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 62 comments

• 0 Temperature is all you need!

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 49 comments

• The missing conversation: Is GPT-OSS by OpenAI a good architecture?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 50 comments

• When LLMs don’t change stuff you don’t want changed

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 8 comments

• Let’s talk about models you believed are more Hyped than Hot

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 21 comments

• OpenAI’s announcement of their new Open Weights (Probably)

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 16 comments

• Speculative Decoding and Quantization ... I'm probably not going anywhere near what you think...

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 9 comments

• Has anyone tried the new ICONN-1 (an Apache licensed model)

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 33 comments

• Llama 4 - Scout: best quantization resource and comparison to Llama 3.3

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 14 comments

• Has anyone gotten featherless-ai’s Qwerky-QwQ-32B running locally?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 20 comments

• How do you edit writing with LLMs: what editor are you using?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 12 comments

• Haters gotta hate, daters gotta date? Llama 4 - a hopeful perspective

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 6 comments

• What software do you use to interact with local large language models and why?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 134 comments

• Before I try... has someone already done it? (Ordered Retrieval Content - ORC)

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 6 comments

• Best solution for a multimodal vision task - OCR

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 19 comments

• Qwen vs OpenAI

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 2 comments

• Kokoro - I'm not the brightest crayon in the box...

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 21 comments

• Grok 2 open source coming?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 11 comments

• Is this guy AI?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 0 comments

• Could this eliminate Qwen’s tendency to slip out of English

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 19 comments

• Multi-Part Prompt with multi-step instructions

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 2 comments

• Would you rather…

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 14 comments

• What secret sauce do we have… and can we share it?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 3 comments

• Best Rag option... that isn't Ollama?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 19 comments

• Why is movement on Yi 1.5 so lackluster?

  Posted by silenceimpaired@reddit | LocalLLaMA | View on Reddit | 32 comments