llama.cpp

The latest AI and IT news about llama.cpp, newest first, with key points summarized by DevPick's AI.

Latest articles and summaries (6)

PC Watchby 劉 尭10/15/2026
AI summary

Hugging Face released 'Llama Windows', a free app allowing users to run local LLMs via the Windows 11 taskbar without commands.

  • •A new app enables running local LLMs on Windows without command-line input.
  • •Hugging Face released 'Llama Windows' on October 13, US time.
  • •The app supports RTX Spark from day one and is available for free.
#Generative AI Tools & Use#LLM / Model Trends
Hacker NewsHN submitter: khalidkhair10/13/2026
AI summary

Overview of Janas, a tool that executes large LLMs exceeding RAM capacity via SSD streaming, achieving faster inference than llama.cpp.

  • •Qwen3-Next-80B (48GB) was processed at 37 tok/s on a 32GB laptop, significantly outperforming llama.cpp's 2.8 tok/s.
  • •All 13 models compared in Q4_K_M quantization format recorded faster read speeds than llama.cpp.
  • •Specific performance figures were presented: 24 tok/s without drafting and 62 tok/s for reading.
#LLM / Model Trends#Semiconductors & Infrastructure
gihyo.jp10/8/2026
AI summary

Microsoft experimentally integrated the llama.cpp LLM inference engine into Windows ML, its local AI inference foundation for Windows.

  • •Microsoft announced the integration of llama.cpp into Windows ML on October 7, 2026.
  • •llama.cpp is an inference engine used to run large language models (LLMs).
  • •This integration is provided as experimental support.
#LLM / Model Trends#Semiconductors & Infrastructure
PC Watchby 劉 尭10/6/2026
AI summary

llama.cpp v0.6.0 was released, adding support for the new GLM-5.3-Flash model and Metal-based acceleration.

  • •ggml-org released llama.cpp v0.6.0 on October 5, which supports the new model 'GLM-5.3-Flash'.
  • •The software is available for free download from GitHub and uses the MIT license.
  • •Metal-based acceleration features were also added, further evolving local LLM execution environments.
#LLM / Model Trends#Generative AI Tools & Use
Hacker NewsHN submitter: anerli10/1/2026
AI summary

エージェント向けにハードウェア上で最大限の速度で動作する自己最適化推論エンジン「Magnitude」が発表され、llama.cppより最大2倍高速であるとされている

  • •Mac、Linux、Windowsの任意のハードウェア上で動作する推論エンジンである
  • •既存の推論エンジンであるllama.cppと比較して最大2倍の速度向上を実現している
  • •開発者は過去に4,000以上のGitHubスターと100,000以上のダウンロードを獲得したオープンソースブラウザエージェントを開発した経験を持つ
#Semiconductors & Infrastructure#AI Agents
Hacker NewsHN submitter: pptadversary9/27/2026
AI summary

llama.cppのn-gramキャッシュに対する4つの変更により、ドラフティング速度が最大41.6倍に向上し、メモリ使用量も削減された

  • •n-gramキャッシュへの4つの変更により、ドラフティング処理が最大41.6倍高速化された
  • •静的キャッシュの読み込み速度が最大23.5倍向上した
  • •ピークメモリ使用量が最大2.65倍低下した
#LLM / Model Trends#Semiconductors & Infrastructure