Claude Opus 4.6 Released

Anthropic upgrades its smartest model with Claude Opus 4.6, industry-leading across agentic coding, computer use, tool use, search and finance.

Anthropic announced Claude Opus 4.6 on February 5, 2026, available on claude.ai, the API (claude-opus-4-6) and major cloud platforms at unchanged pricing of $5/$25 per million tokens^[1][2]
Claude Code gained agent teams as a research preview, and Claude in PowerPoint arrived in research preview^[1]
The Opus product page dates it February 5, 2026 and describes it as Anthropic's most capable model at the time^[2]

Opus 4.6 improves on its predecessor's coding: it plans more carefully, sustains agentic tasks for longer, operates more reliably in larger codebases and has better code review and debugging skills to catch its own mistakes; it also applies these to financial analyses, research, documents, spreadsheets and presentations^[1]
First Opus-class model with a 1M token context window (beta, Claude Developer Platform only), with premium pricing of $10/$37.50 above 200k prompt tokens and outputs up to 128k tokens^[1]
Results: highest score on Terminal-Bench 2.0, leads on Humanity's Last Exam, outperforms GPT-5.2 by about 144 Elo on GDPval-AA, and scores 76% on MRCR v2 8-needle 1M versus Sonnet 4.5's 18.5%^[1]
API features: adaptive thinking, four effort levels (low, medium, high default, max) and context compaction in beta^[1]
Safety: as well-aligned as Opus 4.5 with the lowest over-refusal rate of recent Claude models^[1]

参考资料 2可信度 90%

第一项始终是图钉的来源。总体可信度是各份资料对上方所用开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半。

[1]95%
Claude Opus 4.6
anthropic.com· 发布于 2026年9月19日· 开始 2026年2月5日· 占评分 50%
Anthropic's^[2] own announcement, dated the day the model became available.
[2]85%
Claude Opus \ Anthropic
anthropic.com· 添加于 2026年9月19日· 开始 2026年2月5日· 占评分 50%
Anthropic's^[1] Claude Opus page lists this announcement and its date.

有遗漏或错误吗？用你自己的话说明：能佐证此图钉的链接、不同的开始或结束日期及理由，或缺失、有误的信息。AI 会对照此图钉的来源进行核实，搜索更好的来源，并添加任何支持你说法的页面。图钉自身的来源仍然最重要。