Summary (TL;DR)
Anthropic released Claude Opus 4.6, its smartest model yet, with major improvements in coding, planning, and long-context handling. It features a 1M token context window in beta, 128k output tokens, and new controls like adaptive thinking and effort levels. The model leads on benchmarks including Terminal-Bench 2.0, Humanity's Last Exam, and BrowseComp. Safety evaluations show low misaligned behavior and over-refusal rates. Product updates include agent teams in Claude Code, context compaction for long-running tasks, and better integration with Excel and PowerPoint (research preview). Pricing is unchanged at $5/$25 per million tokens.
Public access expired
Save this link to your readplace queue and read every link without expiration.
Save to My Queue