Nvidia, AMD, Intel, Google, Meta and four others are founding members of the SGLang Consortium. The group gives the industry's biggest rival chipmakers a seat at the table where an open-source inference engine plans its roadmap.
Liquid AI's 74M-parameter streaming audio codec encodes at 361x real time on a single CPU thread, beating Mimi, X-Codec 2, and NanoCodec on speech generation quality.
Google rolls out Gemini 4 Hydrogen globally as its daily driver for agentic workflows, available free via auto model selection and across every product tier.
A new benchmark shows LLMs that refuse biased prompts when identity is stated directly still produce stereotypes when identity is hinted at through cultural cues.
Parallel Web Systems is handing out free API credits to early-stage founders through YC, a16z speedrun, Neo, Entrepreneurs First, Founders Inc and Null Fellows.
Helion now picks fast GPU kernel configs by analyzing IR at compile time, skipping hours of autotuning while beating handwritten Triton and CUDA baselines.
Hark's Handoff² computer-use agent completes web tasks twice as fast as the original while closing the accuracy gap with frontier models to within 5 points.
Salesforce's new 4.6B-active-parameter speech recognition model tops the open weights leaderboard with a 2.3% word error rate, beating Voxtral Small and Nemotron 3.5.
A new gradient inversion attack dissects GANs layer by layer to reconstruct private client images from federated learning gradients at pixel-level fidelity.
Baidu launched a neutral marketplace for AI model services with 100+ models processing 700B tokens daily, aiming to standardize how enterprises buy inference.