AMD Introduces ROCm Hyperloom for Automated Inference Optimization
AMD announced ROCm Hyperloom, an open-source tool that reduces inference workload optimization time from weeks to hours, supporting AMD Instinct GPUs.
Software
490 articles
The tools, frameworks, and platforms developers use to build with AI — coding assistants, agent frameworks, APIs, and developer tooling, with the updates that change how you ship.
LATEST NEWS
-
-
Runway Launches Media Router for Generative Media Models
Runway released its Media Router on July 23, 2026, offering developers API access to a growing list of image, video, and audio models, including its own.
-
AMD AI Workbench Lets Users Deploy Custom Models
AMD AI Workbench enables users to deploy custom models directly through its GUI, supporting Hugging Face repositories and proprietary models.
-
AMD Introduces Kimi-K2.5-MXFP4 Model on MI355X with ATOM
AMD demonstrates how to serve the Kimi-K2.5-MXFP4 model on MI355X using ATOM, enabling optimized FP4 execution without manual configuration.
-
Amazon Introduces Agentic Retrieval for Bedrock Managed Knowledge Bases
Amazon Bedrock now supports agentic retrieval, which handles complex multi-part questions by iteratively planning and retrieving information, improving accuracy for analysts.
-
Amazon Bedrock AgentCore Detects Silent Agent Failures
Amazon Bedrock AgentCore helps identify silent agent failures that pass health checks but cause customer issues, with 99% completion rates and zero error spikes.
-
Amazon Quick Uses Highcharts for Multi-Region Visualizations
Amazon Quick Sight now supports Highcharts for multi-region carrier performance dashboards, enabling unified visualizations without moving raw data across borders.
-
Jefferies Uses AI to Streamline Trading Operations
Jefferies implemented an AI trade assistant to improve real-time data analysis for traders, reducing reliance on IT and experts.
-
AWS Introduces Blueprint for Evaluating AI Agents Using Strands and AgentCore
AWS and Motorway developed a production-ready blueprint to evaluate AI agents, reducing incorrect results from 1 in 8 to 1 in 50.
-
AMD Introduces Radeon AI Support for Enterprise AI Stack
AMD announced technical preview support for Radeon AI PRO R9700 and PRO W7900 GPUs in its enterprise AI reference stack, including the AMD GPU Operator and AIMs.
-
AMD Launches ROCm Infinity Context for Distributed Inference
AMD introduces ROCm AIC, a new KV cache storage tier, to address growing bottlenecks in AI inference. The solution supports up to 10 million token contexts and reduces HBM waste by 600 GB per request.
-
AMD Introduces Spur GPU Scheduler for HPC and AI Workloads
AMD launches Spur, a GPU-first job scheduler, to address challenges in managing GPU clusters for AI and HPC workloads. The tool supports up to 64 MI300X GPUs across 8 nodes.