PANews reported on May 28th that Ethereum co-founder Vitalik Buterin, in response to a discussion, stated that running large models locally should be designed around several standardized "tiers" in terms of memory and VRAM configurations, with 96GB being one of the current feasible upper limits for local deployment. He mentioned that the RTX 6000 has 96GB of VRAM, and high-end MacBooks and some AMD laptops have a unified 128GB of RAM, requiring approximately 32GB to be reserved for other tasks. Vitalik suggested that the entire local LLM and GPU ecosystem should be designed around a few capacity standards such as 2GB, 8GB, 24GB, 96GB, 256GB, and 1TB to unify hardware and model design, thereby improving deployment efficiency and compatibility.