Pinchflat

192GB Framework Desktop: Local AI on Gorgon Halo

Media Tasks
Actions

Raw Attributes

Source: Framework
  • nfo_filepath:
  • duration_seconds: 928
  • prevent_culling: false
  • playlist_index: 0
  • media_filepath: /downloads/Framework/2026-09-30 192GB Framework Desktop： Local AI on Gorgon Halo/192GB Framework Desktop： Local AI on Gorgon Halo [Z26kN5VfyGA].mp4
  • description: The 192GB Framework Desktop uses AMD’s Ryzen AI Max+ PRO 495, codenamed Gorgon Halo. This episode looks at what its extra 64GB of unified memory enables beyond a 128GB Strix Halo system: keeping larger models resident without SSD streaming or a cluster, and running two local models together. Some of the models that can be run on this system are DeepSeek V4.1 Flash, MiMo V2.6 Flash RL and GLM-5.3 Flash. The video covers their performance across growing context depths, shows DeepSeek V4.1 Flash Q2 running through DwarfStar, and demonstrates a two-model workflow in which Qwen3.8-Flash-Next can ask DeepSeek V4 Flash for help. It also covers Framework’s open-ended PCIe x4 slot for longer GPU or network-card connectors. Timestamps 00:00 - Introduction 01:31 - Gorgon Halo hardware 02:28 - Models and performance 10:37 - Running multiple models 12:47 - Open-ended PCIe slot 14:29 - Conclusions Links & Resources Hardware and Framework Framework Desktop: https://community.frame.work/t/introducing-the-framework-desktop/65008 AMD Ryzen AI Max+ PRO 495: https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-plus-pro-495.html Framework guide to choosing a Desktop for local AI: https://frame.work/gb/en/blog/choosing-a-framwork-desktop-for-local-ai Models and quantizations DeepSeek V4.1 Flash: https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash antirez’s DeepSeek V4.1 Flash Q2 GGUF: https://huggingface.co/antirez/deepseek-v4.1-flash-gguf DeepSeek V4.1 technical report: https://arxiv.org/abs/2609.19969 GLM-5.3 Flash: https://huggingface.co/zai-org/GLM-5.3-Flash antirez’s GLM-5.3 Flash Q2 GGUF: https://huggingface.co/antirez/glm-5.3-flash-gguf/tree/main GLM-5.3 Flash mixed Q2/Q4 GGUF: https://huggingface.co/kyuz0/GLM-5.3-Flash-Mixed-Q2-Q4-GGUF/tree/main MiMo V2.6 Flash RL: https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL Qwen3.8-Flash-Next: https://huggingface.co/Qwen/Qwen3.8-Flash-Next Engines and setup AI Toolbox Cockpit: https://github.com/kyuz0/ai-toolbox-cockpit DwarfStar: https://dwarfstar.sh/ llama.cpp: https://github.com/ggml-org/llama.cpp Halogen Flash Server: https://github.com/peonist-ai/halogen-flash-server Adobe second-o-pi-nion: https://github.com/adobe/second-o-pi-nion pi subagents package: https://pi.dev/packages/pi-subagents?name=subagent Speed benchmarks: Local LLM speed benchmarks: https://local-llm-benchmarks.dev/ Related work and videos Qwen3.8-Flash-Next, Halogen and performance: https://www.youtube.com/watch?v=Nm_zN6RQ_eE Low-latency Strix Halo cluster with RDMA and Intel E810: https://www.youtube.com/watch?v=nnB8a3OHS2E Lucebox heterogeneous inference: https://www.lucebox.com/engine Intel E810 network adapters: https://www.intel.com/content/www/us/en/products/details/ethernet/800-network-adapters/e810-network-adapters.html Community and support Strix Halo toolboxes and tutorials: https://strix-halo-toolboxes.com Strix Halo Discord community: https://discord.gg/pnPRyucNrG
  • title: 192GB Framework Desktop: Local AI on Gorgon Halo
  • short_form_content: false
  • id: 1577495
  • subtitle_filepaths: en/downloads/Framework/2026-09-30 192GB Framework Desktop： Local AI on Gorgon Halo/192GB Framework Desktop： Local AI on Gorgon Halo [Z26kN5VfyGA].en.srt
  • livestream: false
  • last_error:
  • media_downloaded_at: 2026-10-04T18:14:36Z
  • metadata_filepath: /downloads/Framework/2026-09-30 192GB Framework Desktop： Local AI on Gorgon Halo/192GB Framework Desktop： Local AI on Gorgon Halo [Z26kN5VfyGA].info.json
  • predicted_media_filepath: /downloads/Framework/2026-09-30 192GB Framework Desktop： Local AI on Gorgon Halo/192GB Framework Desktop： Local AI on Gorgon Halo [Z26kN5VfyGA].mp4
  • uploaded_at: 2026-09-30T14:55:35Z
  • matching_search_term:
  • updated_at: 2026-10-04T18:16:53Z
  • prevent_download: false
  • thumbnail_filepath: /downloads/Framework/2026-09-30 192GB Framework Desktop： Local AI on Gorgon Halo/192GB Framework Desktop： Local AI on Gorgon Halo [Z26kN5VfyGA]-thumb.jpg
  • inserted_at: 2026-09-30T15:00:43Z
  • media_id: Z26kN5VfyGA
  • culled_at:
  • media_size_bytes: 42011674
  • original_url: https://www.youtube.com/watch?v=Z26kN5VfyGA
  • media_redownloaded_at:
  • source_id: 4
  • uuid: 9e75149d-6cab-4f0e-a213-e819628beade
  • upload_date_index: 99
Worker
State
Scheduled At
Pinchflat.Downloading.MediaDownloadWorker completed