Tag
3 stories · 0 tools
OpenAI and Broadcom have unveiled Jalapeño, a custom AI accelerator designed specifically for large language model inference. Early testing indicates the chip delivers substantially better performance per watt than current state-of-the-art hardware, with deployment expected to begin at gigawatt scale by the end of 2026.
OpenAI has begun rolling out a significantly more capable and compute-efficient memory architecture—built on a technique called 'dreaming'—to ChatGPT Plus and Pro subscribers in the US. The update promises a more scalable system for synthesizing conversation memory, including a new memory summary feature for users.
Hugging Face's blog listing showcases recent community and official articles covering open-source humanoid robotics, audio LLMs, transformer inference optimization, and agent benchmarks. The posts span topics from fine-tuning video generation models to local browser deployment and reinforcement learning post-training techniques.