Models & Labs
llama.cpp b10043 release supports CUDA Virtual Devices
llama.cpp Releaseshigh confidence
Why it matters
- →Support for CUDA Virtual Devices enhances GPU resource management.
- →Disabling NCCL path optimizes performance for virtual device configurations.
- →The update improves testing and deployment processes with new CI jobs.
The b10043 release of llama.cpp brings support for CUDA Virtual Devices, allowing for more efficient GPU resource management. This update disables the NCCL path when virtual devices are used, optimizing performance for these configurations. Additionally, the release includes a code refactor and introduces GPUx2 server CI jobs to enhance testing capabilities. While no new model architectures are introduced, the update broadens the platform's adaptability across different operating systems, benefiting developers working with varied hardware.
Read originalMore from llama.cpp Releases
Open Sourcemodels
llama.cpp b10045 Release Expands Platform Support
The b10045 release of llama.cpp focuses on broadening its platform compatibility, though it doesn't introduce major new features. This update notably includes Vulkan support for Ubuntu and Windows, alongside ROCm 7.2 for Ubuntu, enhancing GPU utilization options for developers. While KleidiAI support for macOS Apple Silicon remains disabled, the release still covers a wide array of operating systems and architectures, offering developers increased flexibility in deployment. This update solidifies llama.cpp's position as a versatile inference runtime across multiple systems, rather than delivering groundbreaking changes.
llama.cpp Releases
Open Sourcemodels
llama.cpp b10046 Release Expands Platform Support
The b10046 release of llama.cpp continues to broaden its platform compatibility, making it an adaptable tool for developers working across different systems. This update notably includes support for Ubuntu with ROCm 7.2, which enhances performance for AMD GPU users. Windows users gain from the inclusion of CUDA 12 and 13 DLLs, ensuring they can leverage the latest NVIDIA technologies. While macOS Apple Silicon support remains strong, the KleidiAI feature is temporarily disabled. This release reflects llama.cpp's ongoing effort to be a comprehensive inference runtime across a wide range of hardware configurations.
llama.cpp Releases
Open Sourcemodels
llama.cpp b10047 Release Expands Platform Support
The b10047 release of llama.cpp marks another step in its mission to support diverse hardware environments, now extending compatibility to platforms like macOS, Linux, and Windows. This update brings Vulkan support to Ubuntu and Windows, enhancing their graphics processing capabilities. The addition of ROCm 7.2 for Ubuntu x64 is a significant move for AMD GPU users, offering improved performance. While the release doesn't feature new models, it continues to support CUDA for NVIDIA users, ensuring robust performance across different setups. This version focuses on making llama.cpp a more versatile tool for developers working with various hardware configurations.
llama.cpp Releases
More in Models & Labs
© Hugging Face BlogModels & Labsmodels
NVIDIA Nemotron 3 Embed Tops RTEB Leaderboard
NVIDIA's Nemotron 3 Embed models have set a new benchmark in retrieval quality, with the 8B model ranking #1 on the RTEB leaderboard. This collection of embedding models is designed for production-scale retrieval tasks, offering open weights and datasets for customization. The models support multilingual and code retrieval, and are optimized for high-throughput deployment with NVIDIA's NVFP4 technology. This release provides developers with powerful tools for efficient and accurate retrieval, enhancing capabilities in agentic retrieval and reducing operational costs.
Hugging Face Blog
© The Verge AIModels & Labsproductivity
Google renames NotebookLM to Gemini Notebook
Google's AI note-taking app is getting a new identity as it transitions from NotebookLM to Gemini Notebook. This rebranding aligns with deeper integration into Google's broader AI ecosystem, including Gemini and Google Search. The app, initially introduced as Project Tailwind, has evolved with features like AI-generated podcasts and TikTok-style clips. Now, it also offers a secure cloud computing feature for code execution, initially available to select business customers. This move signifies Google's commitment to enhancing AI-driven productivity tools.
The Verge AI
© TechCrunch AIModels & Labsmodels
Moonshot's Kimi 3 to Rival Anthropic's Opus 4.8
Moonshot AI is on the verge of releasing its Kimi 3 model, which is poised to compete directly with Anthropic's Opus 4.8 in terms of performance. Expected to be the largest open-weight AI model from China, Kimi 3 boasts an impressive parameter count between 2 trillion and 3 trillion. This move signifies a shift in the AI landscape, where open-source models are increasingly challenging the supremacy of closed-source counterparts from companies like OpenAI and Anthropic. As the industry grapples with the costs and data privacy issues of proprietary models, Moonshot's Kimi 3 emerges as a viable alternative for businesses seeking robust AI solutions without the hefty price tag.
TechCrunch AI