Tag: llm
All the articles with the tag "llm".
-
DeepSeek V4 Is the Boring Kind of Terrifying: Cheap, Huge, and Almost There
DeepSeek V4 previewed on April 24 with 1.6 trillion parameters, 1 million token context, open weights, and pricing that attacks frontier AI economics directly.
-
Google Just Made Every AI Model 6x Cheaper to Run. Memory Chip Stocks Crashed.
TurboQuant compresses LLM memory from 16 bits to 3 bits with zero accuracy loss. 6x less memory, 8x faster inference. And the stock market panicked, because apparently nobody learned from the DeepSeek episode.
-
Llama 4 Scout Has a 10 Million Token Context Window. Let That Sink In.
Meta released Llama 4 with a 10 million token context window. That is roughly 20 full novels. The implications for how we use AI are enormous and nobody is talking about the right ones.
-
One Year After DeepSeek R1: What Actually Changed?
A year ago, a Chinese lab nobody was watching released a model that crashed NVIDIA's stock and challenged every assumption about how much compute you actually need. I was watching them before the explosion. Here is what I think really happened.