AiPulse — Issue No. 38

This partnership enables users to use local AI agents at a lower cost.

OpenAI demonstrated through InferenceX benchmark tests that the Jalapeño chip, developed jointly with Broadcom, offers lower latency and higher energy efficiency in inference tasks compared to Nvidia's GB200 and GB300 superchips.

IBM has designed a dual-architecture processor that enables companies to run Arm-based cloud and artificial intelligence software directly without having to migrate their critical workloads from existing mainframe systems.

Intel announced three new architectures incorporating Xeon, GPU, and mobile processors, adopting a heterogeneous hardware approach to accommodate the power, cost, and deployment constraints of agentic AI workloads spanning from the cloud to the network edge.

Apple's launch of the M5 Ultra, which combines four chips using its 2-nanometer manufacturing process and UltraFusion technology, to enhance local artificial intelligence processing capabilities and the performance of Mac devices, stands out as a significant development in the industry.

The controversy sparked in the legal sector over data privacy and geopolitical independence stems from the fact that Harvey's Tenet model, fine-tuned by the OpenAI-backed Harvey on legal data, was built upon the open-source Kimi K3 weights of China's Moonshot AI.

According to TeamT5's analysis, groups linked to the Chinese government have scaled up their operations by automating routine cyberattack tasks, taking advantage of DeepSeek's low security barriers.

Should You Use Open Source Large Language Models?

You must make your decision by evaluating the differences between large-scale adopted models and those with accessibility and customizability.

Y Combinator-backed Nori Robotics aims to differentiate itself from expensive, teleoperation-dependent competitors by offering a low-cost household robot that users can code and train themselves, featuring 3D-printed parts.