AI
Models, agents, chips and the companies racing to ship them.
5 stories
-
OpenAI, Google and Microsoft back one robots.txt line to opt out of AI training
A shared robots.txt token lets a website refuse AI training once for several providers. Support arrives in stages through 2027, and the draft behind it is not yet a standard.
-
OpenAI releases GPT-6.1 Terra, a cheaper mid-tier model aimed at computer use
GPT-6.1 Terra costs half as much as GPT-6.1 Sol and targets desktop automation. OpenAI's own safety addendum rates it Critical for cybersecurity.
-
Claude Haiku 5.5 costs 90% less per token, but a new tokenizer eats part of the saving
Anthropic's Haiku 5.5, released October 7, is priced at $0.10 and $0.50 per million tokens under 100,000 tokens. A tokenizer change and a price cliff at 100K tokens complicate the headline.
-
Claude Opus 5.5 cuts list prices 20% and claims Fable 5.1 performance on most work
Anthropic released Claude Opus 5.5 on September 22 at $4 and $20 per million tokens. The benchmark wins are large, but Anthropic itself says the gap to Fable 5.1 is smaller in real use.
-
OpenAI's GPT-6 Astra is the first model rated Critical for cyber, and it launches gated
OpenAI unveiled GPT-6 Astra on September 3 to enterprise customers with Daybreak access first. It reached the Critical cyber threshold, so the public version refuses advanced cyber tasks.
