Claude Opus 4.6 (Anthropic)

Overchat's AI Hub page confirms Anthropic released Claude Opus 4.6 on February 5, 2026, as a major upgrade to its flagship model with a 1M token context window (up from 200K), adaptive reasoning, and improved coding. It reports benchmark highs including Terminal-Bench 2.0 at 65.4% and Humanity's Last Exam at 53.1%, wit Additional features highlighted include Agent Teams in preview for parallel multi-instance work, Context Compaction for server-side summarization when nearing limits, and a doubled 128K max output, with early users reporting gains on legal documents, writing quality, and fewer predictive phrases. The page also flags th

Caylent reports that Anthropic's Claude Opus 4.6 launched on Amazon Bedrock with substantive capability upgrades beyond a simple version bump. The model is described as Anthropic's most intelligent model to date, topping the Artificial Analysis Intelligence Index, leading on agent tasks, terminal coding, and research b The post details a 1M token context window in preview (previously experimented with on Sonnet 4.5), a new adaptive thinking system with four reasoning-effort levels, and a 128K output token limit for generating full documents in one response. A compaction API in public beta lets developers compact conversation context