createthiscom

• PC Build: Run Deepseek-V3-0324:671b-Q8 Locally 6-8 tok/s

  Posted by createthiscom@reddit | LocalLLaMA | View on Reddit | 173 comments

• native tool calling support for DeepSeek V3.1 just merged in llama.cpp

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

• The dangers of local LLMs: Sleeper Agents

  Posted by createthiscom@reddit | LocalLLaMA | View on Reddit | 24 comments

• samsung 9100 pro 4tb vs WD_BLACK 8TB SN850X for llama.cpp

  Posted by createthiscom@reddit | LocalLLaMA | View on Reddit | 25 comments

• What's your favorite local model for C#?

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

• large model ( 547G ) load time - llama.cpp

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

• This is awkward

  Posted by createthiscom@reddit | LocalLLaMA | View on Reddit | 188 comments

• more than 131k context on a single GPU - llama.cpp

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

• Why does HF not show total size for directories?

  Posted by createthiscom@reddit | LocalLLaMA | View on Reddit | 4 comments

• DIY LLM inference engine learning

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

• Does your AI need help writing unified diffs?

  Posted by createthiscom@reddit | LocalLLaMA | View on Reddit | 5 comments

• ktransformers: DeepSeek_V3_0324:671b-Q4_K_M - 14 tok/s - Open Hands AI

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