Sakhanda Wire
NVDA $153.77 -1.52% MSFT $504.91 -2.53% GOOGL $281.52 -0.78% META $620.99 -0.87% AMZN $256.80 -0.37%

MarkTechPost

396 stories · all sources

MarkTechPost

Google releases open-source TPU Sync for KV cache transfers

Google Cloud has released TPU Sync, an Apache-2.0 C++ library for moving KV caches and model weights across Cloud TPU memory tiers. It treats TPU HBM, pinned host DRAM, shared storage and the data-center network as a single hierarchy, using asynchronous DMA while preserving the tensors’ TPU tile layout. Python wheels are available for PyTorch and JAX.

MarkTechPost

Google launches Gemini 4 Hydrogen for faster agentic coding

Google has released Gemini 4 Hydrogen, a closed-weights model from Google DeepMind aimed at coding and everyday agent workflows. It is available through the Gemini API and Google products, with no downloadable weights or self-hosting option. The model provides a unified context window of 1,048,576 tokens for input, reasoning and output, and accepts text, images, audio, video, PDFs, tables and charts. Its output is text, with a synchronous limit of 65,536 tokens.

MarkTechPost

Liquid AI Releases 74M-Parameter Streaming Audio Codec

Liquid AI has released LiquidCodec, a streaming speech tokenizer designed for the company’s LFM3-Audio-2B and LFM3-Omni-2B models. The codec has about 74 million parameters, including 73.6 million listed on its model card. It processes 24 kHz audio at 12.5 Hz, producing 12 discrete token streams, with each 80-millisecond frame represented by 12 tokens.

MarkTechPost

CopilotKit releases self-hostable OpenBot 1.0 with policy controls

CopilotKit has released OpenBot 1.0, an MIT-licensed workspace for running AI agents under a company’s own control. It can be used locally or deployed through a Helm chart on AWS EKS, Azure AKS, Google Cloud GKE, or another Kubernetes cluster. Teams provide their own agents, models and infrastructure rather than using a fixed vendor environment.

MarkTechPost

AWS releases a free local emulator for more than 200 services

AWS has released AWS Emulator, a free, CC0-licensed program that runs locally and provides an endpoint compatible with AWS APIs. Developers can direct the AWS CLI or SDKs to it without changing application code, using a laptop, development server or CI environment. No AWS account or service fee is required. Resources are held in memory and disappear when the emulator stops.

MarkTechPost

IBM unveils Nighthawk r3 quantum processor with 520 programmable qubits

IBM announced Nighthawk r3, a superconducting quantum processor designed to perform 10,000 quantum operations across 520 programmable qubits. The chip contains 2,034 physical qubits and 10 l-couplers, which are intended to support connections between processors. IBM plans to combine two r3 chips into a 1,040-qubit system in 2027; that system is not yet available.

MarkTechPost

Google launches Gemini Decisions API for calibrated multimodal scoring

Google has introduced the Gemini Decisions API, powered by Gemini Decision 1, a model designed for closed-set decisions rather than text generation. It evaluates up to 1,000 questions against one shared input and returns calibrated probabilities. Supported inputs include text, images, audio, video and PDFs, while question formats cover yes/no decisions, selections among 2 to 255 options, and ordered rubric scores.

MarkTechPost

Google DeepMind releases AlphaGenome 2 for cell-specific variant analysis

Google DeepMind has released AlphaGenome 2, a conditional DNA sequence-to-function model designed to predict regulatory activity and genetic variant effects for particular cell states. It processes a 1 Mb (1,048,576 bp) human or mouse DNA window and one condition per call. The condition is a 256-dimensional embedding that can be derived from raw single-cell RNA sequencing counts, allowing the system to represent previously unseen cell states without fine-tuning.

MarkTechPost

GPU Leaderboard compares cloud GPU prices for AI inference

MarkTechPost has launched GPU Leaderboard, a free service that compares cloud GPU prices for AI inference, along with the GPU Leaderboard: Deploy It Chrome extension. The site collects on-demand and spot pricing from providers’ own pricing pages and ranks 201 live offers from 24 providers by cost per GPU-hour.

MarkTechPost

MBZUAI releases open PAN 2 world model for physical AI

The Institute of Foundation Models (IFM) at MBZUAI has released PAN 2, an open world model designed to simulate how a scene changes after a specified action. It accepts text and, optionally, an initial image, then generates video showing the predicted result. The model, weights, training code and inference code are available under the Apache 2.0 license.

MarkTechPost

Live 3 brings full-duplex timing to voice agents

Fish Audio has released Live 3, a full-duplex voice model designed for real-time voice agents. It continues listening while speaking, allowing callers to pause, acknowledge with sounds such as “mm-hmm,” or revise what they said without automatically disrupting the interaction. The model is intended to handle conversational timing within the speech system rather than relying mainly on separately configured silence rules and thresholds.

MarkTechPost

Meta FAIR releases e-means clustering algorithm in FAISS 1.16.0

Meta FAIR researchers have released e-means, a mini-batch replacement for Lloyd’s k-means algorithm. The method is available in FAISS 1.16.0 under faiss.contrib.e_means and supports CPU or GPU execution with PyTorch tensors or NumPy arrays. It is intended for clustering tasks used in vector indexes, quantizers and data-curation pipelines.