
DeepSeek Releases Speed Boost for AI Model Inference
Open-sourced optimization techniques could make running large language models significantly faster and cheaper for developers.
Launches, releases, and hands-on coverage of the AI tools you'll actually use, coding assistants, agents, creative tools, and the infrastructure underneath.

Open-sourced optimization techniques could make running large language models significantly faster and cheaper for developers.

The AI firm wins government approval for its advanced model with select partners, though broader public deployment remains uncertain.

Seoul plans universal drone training for 500,000 troops, reshaping defense strategy through autonomous systems and AI automation.

The platform now lets developers launch optimized language model servers instantly, lowering barriers to AI inference at scale.

New research reveals how combining multiple prediction strategies improves language model accuracy on specialized vocabulary.

A change to how Google handles search history means your uploaded media could feed machine learning systems unless you disable the feature.

The streaming giant's investment in the indie studio reveals how artificial intelligence companies are reshaping Hollywood's creative ecosystem.

Researchers on both sides of the AI competition share deep concerns about catastrophic risks amid accelerating development cycles.

A comprehensive leaderboard measures how well automatic speech recognition systems perform outside the lab, reshaping how the field evaluates progress.

New framework automates optimization steps, letting enterprises train large language models faster without manual tuning.

A proposed Cross-Origin Storage API could let web applications run machine learning models more efficiently across multiple domains.

New toolkit with 24 working examples aims to simplify development of autonomous AI applications on modest hardware.

The open-source platform combines machine learning with human review to ship updates reliably, setting a model for sustainable AI infrastructure maintenance.

New PP-OCRv6 variants scale from 1.5M to 34.5M parameters, bringing affordable text recognition to resource-constrained devices.

Engineers and researchers debate the engineering practices needed to create autonomous AI systems that can be trusted in production environments.

The departure of a prominent researcher signals shifting dynamics in the race to develop safer, more capable AI systems.

A comprehensive searchable database reveals millions of songs powering generative AI, raising questions about artist consent and fair compensation.

The computational chemist known for AlphaFold breakthroughs brings deep expertise in machine learning to the safety-focused AI company.