XMasterrrr

• GPU Memory Math for LLMs (2026 Edition)

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

• AMA Announcement: Nous Research, The Opensource Lab Behind Hermes Agent (Wednesday, 8AM-11AM PST)

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 18 comments

• Stop Wasting Your Multi-GPU Setup With llama.cpp: Use vLLM or ExLlamaV2 for Tensor Parallelism

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 112 comments

• AMA Announcement: StepFun AI, The Opensource Lab Behind Step-3.5-Flash Model (Thursday, 8AM-11AM PST)

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

• AMA Announcement: MiniMax, The Opensource Lab Behind MiniMax-M2 + Gifts to Our Community (Wednesday, 8AM-11AM PST)

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

• AMA Announcement: MiniMax, The Opensource Lab Behind MiniMax-M2.5 SoTA Model (Friday, 8AM-11AM PST)

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 28 comments

• MiniMax onX: Weights dropping REALLY, REALLY, SOON

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

• AMA Announcement: Moonshot AI, The Opensource Frontier Lab Behind Kimi K2.5 SoTA Model (Wednesday, 8AM-11AM PST)

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

• AMA Announcement: Z.ai, The Opensource Lab Behind GLM-4.7 (Tuesday, 8AM-11AM PST)

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

• AMA With Z.AI, The Lab Behind GLM Models

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 354 comments

• AMA Announcement: Moonshot AI, The Opensource Frontier Lab Behind Kimi K2 Thinking SoTA Model (Monday, 8AM-11AM PST)

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 47 comments

• Home Server Final Boss: 14x RTX 3090 Build

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 291 comments

• AMA Announcement: Prime Intellect — The Open‑Source Distributed Training Lab (Thu, Oct 2 • 10 AM – 1 PM PDT)

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

• Now I need to explain this to her...

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 517 comments

• Our 4th AMA: The LMStudio Team! (Thursday, 11 AM-1 PM PDT)

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

• Our 3rd AMA: Unsloth Team, Creators of the lightning-fast Unsloth fine-tuning library! (Wednesday, 10 AM-1 PM PST)

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 28 comments

• Our 2nd AMA: Hugging Face Science Team, Creators of SmolLM, SmolVLM, and more! (Tomorrow, 8AM-11AM PST)

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

• GLM-4.5 is now leading the Berkeley Function-Calling Leaderboard V4, Beating Opus 4

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

• Launching Our New AMA Series With Z.AI, Creators of GLM (Tomorrow, 9AM-12PM PST)

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 29 comments

• Qwen 2.5 (7B/14B/32B) Finetunes Outperforming Opus 4 & Sonnet 4/3.5 on Out-of-Distribution Tasks with RL --- Code, Weights, Data, and Paper Released

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 29 comments

• DFLoat11 Quantization for Qwen-Image Drops – Run It on 17GB VRAM with CPU Offloading!

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 18 comments

• I Built My Wife a Simple Web App for Image Editing Using Flux Kontext—Now It’s Open Source

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 77 comments

• So You Want to Learn LLMs? Here's the Roadmap

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

• The Scariest Thing In LLMs/AI Isn't the Models or the Math... It's the Names.

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 23 comments

• o3-mini won the poll! We did it guys!

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 245 comments

• TraceBack: A Novel Reverse Reasoning Model for Better and Cheaper Scaling of Synthetic Reasoning Generation

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 17 comments

• TraceBack: Novel Reverse Reasoning for Better and Cheaper Scaling of Synthetic Reasoning Generation

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

• I Live-Streamed DeepSeek R-1 671B-q4 Running w/ KTransformers on Epyc 7713, 512GB RAM, and 14x RTX 3090s

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 108 comments

• DeepSeek silently released their DeepSeek-Coder-V2-Instruct-0724, which ranks #2 on Aider LLM Leaderboard, and it beats DeepSeek V2.5 according to the leaderboard

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 42 comments

• DeepSeek API: Every Request Is A Timeout :(

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 106 comments

• LocalLLaMA Home Server Final Boss: 14x RTX 3090 Build

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

• Petition to auto-delete anything that mentions Matt Shumer, "Reflection", or any link to his Twitter or any affiliated Twitter accounts (Sahil, etc)

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 46 comments

• Serving AI From The Basement — Part II: Unpacking SWE Agentic Framework, MoEs, Batch Inference, and More · Osman's Odyssey: Byte & Build

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

• Serving AI From The Basement - 192GB of VRAM Setup

  Posted by XMasterrrr@reddit | LocalLLaMA | View on Reddit | 72 comments

• Serving AI From The Basement - 192GB of VRAM Setup

  Posted by XMasterrrr@reddit | programming | View on Reddit | 10 comments

• Serving AI From The Basement - Intro to My 192 GB VRAM Setup · Osman's Odyssey

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