Editorial - SiliconNoon (Page 117)

The Stack

Nvidia's Jetson Orin Nano 2 doubles edge inference with 40% lower power draw

Nvidia just refreshed the little computer that powers a huge slice of the world's robots — and the efficiency numbers are the real story. Nvidia announced the Jetson Orin Nano 2, a new entry-level robotics computer that doubles the inference performance of its predecessor while drawing 40% less power in 15-watt mode. The module packs 78 trillion operations per second of AI compute, 8GB of memory and an eight-core Arm CPU into the same compact form factor as the previous generation, with gains c

The Frontier

IBM's open Granite 4.2 models learn agent skills via reinforcement learning

A string of open-weight releases this week keeps pushing what small, deployable models can do. IBM is the latest, rolling out a reasoning-tuned Granite family that trains the bigger variants to actually use tools and code in real sandboxes before shipping them under a permissive Apache 2.0 license. IBM released Granite 4.2, a family of dense, open-weight reasoning models in 3B, 8B, and 30B sizes, its first to be post-trained with reinforcement learning that teaches them agent-style behavior. Ea

The Stack

Perplexity's Portable Computer runs AI agents entirely offline

Perplexity and Nvidia are making the most aggressive push yet to pull AI agents off the cloud: a desktop app where the model, your files, and the work itself never have to leave the machine — and the token meter never moves. Perplexity launched Portable Computer on Monday, a version of its agentic Computer platform that runs entirely on local hardware — starting with Nvidia's DGX Spark desktop machine and Linux PCs with RTX GPUs carrying at least 24GB of VRAM. Built with Nvidia, it bundles the

The Arena

Jalapeño's first benchmarks put OpenAI on Nvidia's turf

OpenAI's decision to design its own inference chip stopped being a rumor and became a competitive threat today. Fresh off a June announcement, the company published the first real measured results for Jalapeño, the accelerator it co-built with Broadcom — and the numbers, if they hold, are the kind that supply-chain analysts and Nvidia's investor-relations team both lose a little sleep over. The headline "work per watt" gains we flagged in our morning brief are worth sitting with, because the dee

The Stack

OpenAI's Jalapeño chip beats Nvidia superchips on power and latency

Two hardware stories today, both about who sets the price of running AI: OpenAI's first custom inference chip just posted benchmark numbers ahead of Nvidia's best, and one of autonomous trucking's quietest operators just bankrolled its expansion. OpenAI's custom Jalapeño chip delivered 1.5–1.9x more AI work per watt than Nvidia's GB200 and GB300 superchips, with 1.7–3.6x lower end-to-end latency, according to results the company presented at the Hot Chips conference on Tuesday. The numbers come

The Frontier

Apple's STARFlow2 makes normalizing flows a rival to diffusion

While most of the industry keeps refining diffusion, Apple's research group is betting on a cleaner architecture for models that understand and generate text and images in one system. Apple's machine learning research group published STARFlow2, a unified multimodal model built on autoregressive normalizing flows — a direct challenge to the diffusion pipelines that power nearly every image generator today. The paper's starting observation is disarmingly simple: an autoregressive normalizing flow

The Stack

Apple's new Mac Studio and Mac mini are built for local AI

No event, no keynote — just a surprise hardware drop with one very clear audience: people who run AI models on their desks. Apple refreshed its entire desktop lineup around local AI inference today, debuting the M6 — its first 2nm chip — in a new Mac mini and the M5 Ultra, its first-ever quad-die design, in a new Mac Studio. The M6 brings a 12-core CPU and 12-core GPU with a Dual 16-core Neural Engine, and Apple claims the world's fastest single-threaded performance. The M5 Ultra goes further:

The Arena

Google launches Gemini Enterprise for Legal to take on Claude in law

The legal AI land grab just got its biggest player yet — Google has stopped dabbling and shipped a full vertical. Google Cloud launched Gemini Enterprise for Legal on Tuesday, a purpose-built agentic AI package that plugs Gemini into the systems lawyers actually use — the company's most serious move yet into enterprise software's most lucrative vertical. Entering preview with launch customers including legal teams at Cleary, Freshfields, Weil, and Williams & Connolly, the offering bundles speci

The Guardrails

UK gains first access to Ukraine's AI battlefield dataset

Ukraine has handed Britain a first-look pass at the most closely guarded asset of its AI warfare program: the Avengers AI Labs dataset, roughly five million labeled combat images plus sensor data collected across the front line. President Volodymyr Zelensky and Prime Minister Andy Burnham signed the agreement in Kyiv, making the UK the first foreign country granted access to a trove that trains models to spot tanks, artillery, drones, and infantry. Three British startups — Bristol's Sintela, Oxf

The Frontier

Accelerated Understanding launches a physics foundation model

AI's frontier-pushing crowd spent the week cleaving to language and agentic models — but one of the field's better-known researchers used Tuesday to bet on a different kind of foundation model: one that predicts how the physical world changes, not how the world talks. Caltech professor Anima Anandkumar, a former Nvidia machine-learning research director, unveiled a startup and a model aimed squarely at enterprise engineering, promising a single system for chip design, weather, and geology. Acce

Quick Hits

Quick Hits — August 25, 2026

A quiet Monday on the model front, with one architecture preview, one new embedding family, and one sovereign-AI shuffle in Europe. Qwen previews its next-generation Qwen4 architecture in an early open-weight drop. Qwen3.8-Flash-Next is a multimodal mixture-of-experts model built on the upcoming Qwen4 architecture, released ahead of the full Qwen4 family so the community can adapt tools and pipelines early. It lands open-source on ModelScope on August 26, plus an FP8 quantized variant. The comm

The Frontier

Claude Opus 5 lifts its test pass rate to 88.6% — and writes 2.3x more code

Code-analysis firm Sonar put Anthropic's Claude Opus 5 through its standard LLM evaluation and found the model passes more of the tests it's given — hitting 88.6% on executable coding tasks versus 82.9% for the Opus 4.8 it replaces — while generating more than twice as much code along the way. The pass-rate gain is the headline. Across 544 HumanEval and MBPP tasks with executable tests, Opus 5 landed at 88.6%, a 5.7-point step up from Opus 4.8's 82.9%. The three Opus releases before it all clus

See all