Claude Opus 4.6 was Anthropic's most capable model until Claude Opus 5.5 arrived, and it is still a strong brain to put behind Magnus. When the work is genuinely hard - a dense contract, a gnarly plan, a long research thread - Opus 5.5 is now the one to reach for, with 4.6 a solid choice.
Launched February 5, 2026 by Anthropic. Opus 4.6 introduced adaptive thinking (the model decides how long to reason before answering), a 1M-token context window in beta, and up to 128K output tokens - at release it posted the highest agentic-coding scores Anthropic had ever shipped.
On the All-inclusive plan an Opus 4.6 reply costs 5 credits - the premium tier, priced like the heavyweight it is.
Give Opus to your strategist agent: goal planning, long-document drafting, contract review, and debate rounds where you want the strongest argument in the room. It runs the full Claude toolset - integrations, live web search, native PDF reading - with the most headroom for multi-step work.