News
What's happening in robotics and AI — curated by the wujec.ai editors.
OpenAI's smartest model now answers 14 times faster — on chips OpenAI did not build
OpenAI announced on 13 August 2026 an early preview of Ultrafast, a new API service tier that runs GPT-5.6 Sol at up to 750 output tokens per second — up to 14 times faster than standard processing. The company has published no price and no general availability date. Access is limited to a selected group of customers, with a sign-up form for everyone else. The engineering is not OpenAI's. Ultrafast runs on Cerebras wafer-scale processors, which hold 44 GB of SRAM on the chip itself and therefore sidestep the memory-bandwidth ceiling that limits token generation on conventional GPU clusters. That is the fact worth pausing on: the most capable model OpenAI sells reaches its highest speed on silicon designed by another company, and neither NVIDIA nor OpenAI's own hardware programme is in the sentence. OpenAI frames the change as a shift in what has to be traded away. Until now, real-time response meant reaching for a smaller or more specialised model; the company's phrase for the alternative is "more useful work per second". The uses it names are ones where the clock is part of the problem — reading logs and recent commits while an outage is still unfolding, assessing transactions while market conditions move, answering a shopper before hesitation turns into an abandoned cart. Internally, OpenAI says research batches that used to run overnight now fit inside a working day as several iterations. One widely repeated comparison should be read with care. The figures showing GPT-5.6 Sol on Ultrafast completing all 2,500 questions of Humanity's Last Exam in 11 hours 11 minutes against 78 hours 27 minutes for Claude Fable 5 come from Cerebras, the hardware supplier, not from an independent evaluator — and they measure wall-clock time for a benchmark run, not accuracy. OpenAI's own claim is narrower and more useful: on GDP-Val, its benchmark of economically valuable knowledge work, it reports a 5.6-fold end-to-end speedup over standard processing with no loss of quality. Ultrafast is a service tier, not a new model. The weights are the same GPT-5.6 Sol already in the catalogue; what changed is where they run and how fast the tokens come out.
GPT-5.6 Sol →OpenAI's first gadget takes shape: a palm-sized speaker with moving parts, over $300, in 2027
Bloomberg reported on 6 August 2026 the most concrete description so far of the hardware OpenAI has been building with Jony Ive's LoveFrom studio: a doughnut-shaped smart speaker about the size of a hockey puck, priced between roughly $300 and $400, and due to go on sale in 2027. According to the report, the device is battery-powered and meant to be carried from room to room with one hand, built from a premium metal, and fitted with speaker grilles and microphones for spoken conversation with ChatGPT. Two details set it apart from the smart speakers that came before it. First, a camera system and additional sensors are said to feed the assistant information about the room it is standing in, so that it responds to what is happening around it rather than only to spoken commands. Second, the product reportedly has moving parts whose stated purpose is to give it a personality — a design decision that puts it closer to a small domestic robot than to a static speaker. OpenAI is said to still plan an unveiling later this year, ahead of the 2027 launch. All of this comes from people described as familiar with the project rather than from OpenAI itself: the company has not published specifications, a launch date or a price, and the details may change before the product is shown. wujec.ai does not maintain a profile of the device and will not open one until OpenAI describes it officially.
AMD goes after Nvidia's grip on robot brains
AMD used its Advancing AI 2026 event on 23 July to enter physical AI in earnest, announcing Kria AI system-on-modules built around the new Ryzen AI Embedded X100 processors, together with a robotics developer platform that pairs the module with a carrier board and evaluation kit. The hardware is aimed squarely at the compute box inside a humanoid or an autonomous machine: up to sixteen Zen 5 cores for real-time control, an RDNA 3.5 integrated GPU, an NPU for inference and, in the top configurations, up to 128 GB of unified memory. AMD quotes more than 8,000 control decisions per second and vision-language-action reasoning below 100 milliseconds, and benchmarks the platform against Nvidia's Jetson line — claiming 3.4x better real-time reliability, 1.6x more free CPU cores and support for 2.3x more concurrent agents than the Jetson T5000. Independent measurements are not yet available; these are vendor numbers. The strategic point is less about any single figure than about choice. Nvidia's Jetson Thor has been close to the default brain for humanoid developers, and AMD is selling an open stack — Linux, ROCm, a hypervisor, PyTorch and ONNX — to robot makers who would rather not be locked to one vendor. The modules are sampling with early customers now, with general availability through ODM partners expected in the fourth quarter of 2026.
1X gives NEO hands that can feel a grip slipping — 25 force-controlled degrees of freedom, IP68, 10,000 units a year
1X has replaced the hands on its NEO humanoid with a tendon-driven design that the company positions as its answer to the hardest part of home robotics: touching things that were not designed for robots. The numbers 1X publishes are unusually specific. The hand carries 25 degrees of freedom — 22 fully actuated in the fingers and palm, three at the wrist — and, crucially, all 25 are natively force-controlled and fully backdrivable. That is the difference between a gripper that closes to a commanded position and one that closes until it feels the right amount of resistance. The drive is quasi-direct: 1X's own tendon system running gear ratios of roughly 5:1 to 15:1, low enough that external forces travel back through the mechanism instead of being absorbed by a gearbox. Strength and delicacy are quoted side by side: 3.5 Nm peak torque at the thumb CMC joint, 2.6 Nm at the finger MCP joints, up to 45 N of distal flexion force and 17.75 Nm at the wrist, with positioning accuracy of ±0.2 mm. Tactile sensing runs across the fingertips and surfaces and measures three things — normal force, contact location and shear. Shear is the one that matters for the headline: it is what lets the robot notice an object starting to slide and tighten its grip before the object falls, rather than after. Two details point at a machine meant to live in a kitchen rather than a lab. The hand is rated IP68 and built from food-safe polymers and components, so it can be washed. And 1X says wrist joints have been proven beyond two million cycles under high loads, with components tested to millions of cycles. The part that is hardest to fake is manufacturing. 1X says hundreds of the hands have already come off a scalable production line, with capacity for 10,000 hands in 2026. "These hands are the culmination of intensive engineering focused on making humanoids truly useful," said chief executive Bernt Børnich. Hands remain the bottleneck for every humanoid programme: they are the component with the worst ratio of degrees of freedom to available space, and the one that wears out first. A vendor quoting cycle-life figures and a production line — rather than a demo video — is the more telling signal here.
NEO →