Two benchmarks, opposite answers, and the one that was wrongI benchmarked two local models on a 2x DGX Spark cluster. My automated suite picked a winner. Driving the same pages in a real browser picked the other one, and the confident suite was the wrong one.
5 min
Shop note
DeepSeek Ditched Nvidia and Utah Let AI Write PrescriptionsDeepSeek V4 runs on Huawei chips without a single Nvidia GPU, Utah lets AI renew prescriptions for real patients, Google's TurboQuant shrinks model memory 6x, and MCP goes open governance.