Overview
- Opus 4.6 adds a one‑million‑token context window in beta, doubles maximum output to 128,000 tokens, and introduces Context Compaction and Adaptive Thinking for more complex tasks.
- Claude Code debuts Agent Teams in a research preview that runs multiple model instances in parallel under a coordinating session, with each instance billed separately.
- Anthropic reports leading results on benchmarks such as Terminal‑Bench 2.0, Humanity’s Last Exam and GDPval‑AA versus both prior Claude models and competitors.
- Pricing is listed at $5 per million input tokens and $25 per million output tokens, with higher rates for very large prompts and a 10% surcharge for U.S.-only inference.
- The company says the model identified more than 500 previously unknown serious open‑source vulnerabilities as finance‑analysis stocks fell, while OpenAI released a competing coder update and its CEO criticized Anthropic.