Models & Labs
Thinking Machines Launches Inkling Multimodal LLM
Hugging Face Bloghigh confidence
Why it matters
- →Inkling's ability to process multiple modalities simultaneously opens new avenues for AI applications.
- →The Mixture-of-Experts architecture allows for efficient use of computational resources.
- →This model sets a new benchmark for multimodal reasoning capabilities in AI.
Thinking Machines has launched Inkling, a new multimodal large language model available on Hugging Face. Inkling boasts nearly 1 trillion parameters and can handle text, image, and audio inputs simultaneously, making it a versatile tool for multimodal reasoning. The model uses a Mixture-of-Experts architecture, which allows for efficient inference by activating only a portion of its parameters at any given time. This release represents a significant advancement in AI, enabling developers to create more sophisticated applications that can understand and reason across different data types.
Read originalMore from Hugging Face Blog
© 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
© Hugging Face BlogMore in Models & Labs
Models & Labsmodels
llama.cpp b10043 release supports CUDA Virtual Devices
The b10043 release of llama.cpp marks a notable enhancement with the addition of CUDA Virtual Devices, which significantly improves GPU resource management. By removing the NCCL path when virtual devices are in use, the update fine-tunes performance for these specific setups. This release also includes a comprehensive code refactor and the implementation of GPUx2 server CI jobs, reflecting a commitment to better testing and deployment processes. While there are no new model architectures, the update enhances the platform's flexibility across various operating systems, making it more adaptable for developers working with a wide range of hardware configurations.
llama.cpp Releases
Models & Labsmodels
Llama.cpp b10051 Release Enhances SME Kernel Dispatch
The latest release of llama.cpp, b10051, addresses a critical issue in kernel dispatch by distinguishing between SME and SME2 capabilities. Previously, the integration treated SME as a single capability, leading to incorrect dispatch on SME(v1)-only hardware due to the use of SME2-specific instructions. This update introduces both build-time and runtime distinctions, ensuring that kernels are dispatched based on actual hardware support. This refinement enhances the accuracy and efficiency of operations on different hardware configurations, marking a significant improvement for developers working with these systems.
