Technology News
545 articles
AI technology breakthroughs, research papers, patents, and technical innovations from leading AI startups.
Sapient trains competitive 1B-parameter foundation model for $1,500, challenging cost assumptions
Sapient researchers have trained a 1-billion-parameter reasoning foundation model from scratch on 40 billion tokens for approximately $1,500, achieving performance that rivals larger 2B-7B parameter m...
Xiaomi launches MiMo-V2.5-Pro-UltraSpeed model achieving 1,000+ tokens/s throughput on general-purpose GPUs
Chinese consumer electronics and AI company Xiaomi has released MiMo-V2.5-Pro-UltraSpeed, a high-speed variant of its flagship MiMo-V2.5-Pro model. The 1-trillion-parameter model supports 1M-token con...
Moore Threads releases MusaCoder, first domestic GPU to complete full AI model training chain
Moore Threads (摩尔线程), a Chinese GPU startup, has announced MusaCoder, which it claims is the first domestic (Chinese) GPU to support the full AI model training chain — from data preprocessing through...
Anthropic releases Claude Fable 5; Microsoft restricts employee use over data retention concerns
Anthropic yesterday released Claude Fable 5, its first Mythos-class frontier model, marking a step-function advance in capability — but the model's accompanying safety architecture has immediately cre...
Cohere open-sources North Mini Code agent, runs on single H100, ranks 8th in speed
Cohere has released an open-weight coding agent called North Mini Code, designed to run on a single NVIDIA H100 GPU. The agent ranks 8th among open-weight models for output speed and produces 3x more...
Apple AI runs on Nvidia chips. At a WWDC 2026 tech talk, Apple disclosed that its Private Cloud Comp...
Apple AI runs on Nvidia chips. At a WWDC 2026 tech talk, Apple disclosed that its Private Cloud Compute infrastructure for Apple Foundational Model relies on Nvidia hardware hosted within Google Cloud...
OpenAI proposes mandatory AI safety assessment framework, diverging from Trump administration's voluntary NSA-led approach
On June 5, 2026, the Trump administration issued an executive order titled "Advancing Frontier AI Innovation and Security," while OpenAI simultaneously released a white paper titled "Democratic Govern...
CXMT achieves DDR5 pricing parity with Samsung, SK Hynix, and Micron, gaining client-market supply edge
Chinese memory chipmaker CXMT has reached pricing parity with global DRAM leaders Samsung, SK Hynix, and Micron for DDR5 memory, while also holding a supply advantage in the client (PC/laptop) market...
AI agents now drive more web traffic than humans globally, says Cloudflare
Cloudflare CEO Matthew Prince announced that for the first time in internet history, AI agents and bots now generate over 57% of all HTTP requests worldwide, surpassing human traffic. The milestone, w...
Stripe builds deterministic payment infrastructure for autonomous AI economy
Stripe Principal Software Engineer Steve Kaliski outlined the company's strategy for enabling AI agents to conduct secure financial transactions at an AI Engineer Europe event. Kaliski introduced Stri...
OpenAI has introduced a new memory system codenamed 'Dream' for its ChatGPT Plus and Pro subscribers...
OpenAI has introduced a new memory system codenamed 'Dream' for its ChatGPT Plus and Pro subscribers. The system aims to improve conversational continuity by enabling the model to retain and recall us...
Walmart builds internal multi-LLM coding agent Code Puppy to escape vendor lock-in
Walmart has developed a proprietary AI coding agent called "Code Puppy" that integrates dozens of large language models from providers including OpenAI, Google, and Anthropic. Created by senior engine...
Uber exhausted its 2025 AI coding tool budget within four months, with roughly 5,000 engineers using...
Uber exhausted its 2025 AI coding tool budget within four months, with roughly 5,000 engineers using agentic coding tools at costs ranging from $150 to as high as $2,000 per engineer per month. CFO An...
ZhiZaiWuJie (智在无界) launches Being-H-Flash, an implicit world model for robots that runs on edge devi...
ZhiZaiWuJie (智在无界) launches Being-H-Flash, an implicit world model for robots that runs on edge devices with as little as 100 TOPS, achieving real-time inference at ~20 FPS. The company claims monthly...
ModelBest (面壁智能) holds 'Open Source Week', releasing five edge-AI technologies across full stack
From May 25 to 29 2026, ModelBest (面壁智能), the Beijing-based edge-AI pioneer, and its OpenBMB community staged a five-day 'Open Source Week' releasing daily one major technology: a 1.58-bit low-bit tra...
XCENA targets AI memory bottleneck with near-memory computing
South Korean chip startup XCENA has developed a computational memory chip called MX1 that uses near-memory computing to address the memory bottleneck in AI inference. The MX1 attaches processing cores...
Cerebras positions itself as the AI compute alternative for buyers seeking non-Nvidia solutions, lev...
Cerebras positions itself as the AI compute alternative for buyers seeking non-Nvidia solutions, leveraging geopolitical tensions to grow its chip business. The company is actively marketing its wafer...
MiniMax M3 launch challenges closed-source frontier with Token Plan pricing and three-in-one capability
Chinese AI lab MiniMax has released the M3 model, a flagship foundation model that combines long-context (1M tokens), native multimodal (text+image from pretraining), and advanced coding capabilities...
**NVIDIA releases Cosmos 3 open physical AI foundation model with hybrid transformer architecture.**
NVIDIA has released Cosmos 3, an open physical AI foundation model featuring a hybrid transformer architecture. The model is designed to advance physical AI — the application of AI to understand and i...
MiniMax releases M3 frontier model for coding agents with one-million-token context
MiniMax, the Shanghai-based AI lab listed on the Hong Kong Stock Exchange, has launched M3, a frontier model designed for coding agents with a one-million-token context window and native multimodal in...