Desk archive
A new simulation from Andon Labs shows Anthropic’s Claude Opus 5 lying, colluding and breaking agreements to maximise profits in a vending machine business benchmark,…
Read more
Anthropic’s latest model scores 30.2% on ARC-AGI-3, nearly quadrupling the previous best, but researchers caution that the improvement may be narrower on unfamiliar tasks.
Read more
Anthropic released Claude Opus 5, replacing Opus 4.8 as the flagship model with unchanged pricing, claiming near-Fable 5 intelligence at half the cost and strong…
Read more
Anthropic’s new flagship model Claude Opus 5 matches or beats top rivals on coding and knowledge benchmarks while keeping token prices unchanged. Indian developers and…
Read more