The AI21 Labs team shared their experience optimizing vLLM – a popular tool for deploying language models that often faces critical errors due to RAM shortages when scaling.
Roblox has presented Cube, its proprietary model for generating three-dimensional scenes. The tool is designed to simplify spatial design and help users create content within the platform faster.
Mistral AI has unveiled Voxtral – a real-time speech transcription model featuring precise speaker separation and a new interactive «sandbox» for audio workflows.
AI: Events
Tencent Open-Sources HPC-Ops Library: How to Accelerate Large Model Inference by 30%
Technical context • Infrastructure
The Chinese company has released a set of optimized operators for working with Large Language Models (LLMs) – promising a noticeable speed boost without altering the architecture.
LG AI Research shared details about K-EXAONE – a multimodal model developed using proprietary technology, specifically tailored for the Korean language and cultural context.
Lab
How to Distribute the «Brain» Among Antennas: A New Architecture for Borderless Networks
Electrical Engineering & System Sciences
When every access point becomes a local coordinator rather than just a repeater, the network runs faster without overloading the data center.
Apple has added autonomous programming capabilities to Xcode – now the AI assistant can independently solve development tasks rather than just completing code.
The company has released an improved version of its model, designed to adapt user interfaces for various languages and cultures.
Anthropic and Apple have reached an agreement: developers can now summon the AI assistant Claude from the code editor – faster and without switching between windows.