The short answer
● Anthropic cut Claude Fable to $8 — Fable 5.2 and Mythos 5.2 shipped at $8/M input, 20% below 5.1, six days before the IPO.
● Meta hit the same $0.10 as everyone else — Muse Ember matches Claude Haiku 5.5 and gpt-6-luna exactly. Price stopped being the differentiator.
● NVIDIA gave away 100,000 hours of video — GR00T 2 and Isaac Lab 3.0 are open, and the training footage came with them.
Three things happened on Monday, and two of them were about price.
Anthropic ships cheaper models the week of its IPO
Claude Fable 5.2 and Mythos 5.2 arrived at $8 per million input tokens and $40 output, down from $10 and $50. Same base model; Fable is open to everyone with cyber and bio safeguards, Mythos lifts those limits for vetted research organisations only. Anthropic prices its listing the week of 16 November, targeting above $1 trillion. Full story.
Meta lands on the industry price for a cheap model
Muse Ember launched at $0.10 per million input tokens — identical to Claude Haiku 5.5 and OpenAI gpt-6-luna. Fourth in the Muse line, handles text, images and audio, and is built to execute what Muse Wonder plans. Full story.
NVIDIA opens GR00T 2 and the data behind it
Announced at CoRL in Austin with 100,000 hours of first-person human video. 40 of 50 tasks cleared at least once untrained, 71% after one demonstration, 95% after five — on tasks NVIDIA selected as learnable. Isaac Lab 3.0 is open source and trains up to 2.5 times faster. Full story.
What we published today