NewDecoded All our Data news
Databricks Says Proteus Wrote Qwen Kernels Up to 5.2 Times Faster Than vLLM
The company's Proteus harness generated shape-specific GPU kernels for Qwen 3.5 122B and said the hard part was checking the drafts, not writing them.
Marco Bianchi
3 min read
News 3 min read
Snowflake Says Iceberg Read Restrictions Put Row and Column Rules in the Catalog
News 3 min read
Google Ships WeatherNext 3 With Hourly Satellite-Grounded Forecasts
News 2 min read
Microsoft Ships MAI-Transcribe-2 at 10 Cents an Hour
News 3 min read
OpenAI Connects Epic Records and Public Health Data to ChatGPT
News 2 min read
Nscale Closes About $3 Billion in Rated Loans for Texas and North Carolina AI Campuses
News 3 min read
EuroHPC Signs the LUMI-AI Supercomputer Contract for Kajaani
News 3 min read
Microsoft Sets November for the Saudi Arabia East Cloud Region
News 2 min read
Metriport Raises $26M to Turn Scattered Health Records Into Point-of-Care Context
News 2 min read
MTN and UAE Partner Form Africa Data Hub for AI-Ready Capacity
News 2 min read
Schwarz Group Plans a 5.6 Billion Euro Data Center in Mecklenburg-Vorpommern
News 3 min read
Emerald AI Raises $150M Series A to Make Data Centers Flex With the Grid
Featured
Gimlet Labs Raises $300 Million Series B for a Multi-Silicon Inference Cloud
News 3 min read
Ant Group's inclusionAI Opens LLaDA-Image and a 4-Step Turbo Twin
News 2 min read
IFM Releases K2 Horizon, a Fully Open Six-Model Fleet
News 3 min read
NVIDIA Agrees to Buy Hugging Face for $12.93 Billion
News 3 min read
Want to advertise your Data, Analytics, or AI here? Reach out!
NewDecoded