Model Evaluation

A comprehensive buying guide for Apple's 2026 M5 and M6 hardware lineup tailored for running massive open-weight Local AI models like Qwen 3.8 and DeepSeek V4. We break down the RAM requirements, footprints, and cost comparisons to help you decide between local hardware and cloud APIs.

Tencent's new Hy-4 Preview model offers a 770B MoE architecture at under $1 per million input tokens. Here is what its Apache 2.0 release and self-optimized training mean for your AI builder stack.

Three numbers, three steps, one pattern: pretraining researcher Jacob Coxon resigns from Anthropic — and alignment lead Evan Hubinger answers within 24 hours with numbers. The debate gains a counting system of case, probability, and trend.

New interpretability work from Anthropic read Claude's internal activations and found evaluation awareness on up to 26% of benchmark problems — even when the model never said so. Here's what it changes for teams that ship on eval scores.