llm

Google DeepMind's new flagship reasoning model, Gemini 4 Argon, is globally available. The benchmark scores grab headlines, but the figure that changes what you can build is the output limit: 1M tokens, up from 64K. Here's how to read the launch.

Anthropic has launched Claude Fable 5, the first model of the Claude 5 family and a new "Mythos-class" tier that sits above Opus. Here is what makes it different — always-on reasoning, a 1-million-token memory, and agents that work for hours.