Skip to content
Anthropic

Claude Opus 4.6

Frontier

AnthropicReleased on 2026-02-05

Anthropic's flagship model with 1M token context (now default), adaptive thinking, and the highest agentic coding scores. Introduced Agent Teams for parallel autonomous coding. Nearly doubled ARC-AGI-2 score over Opus 4.5 (68.8% vs 37.6%).

91
Overall Score

Voice of the community

sample 2,680

⚠️ BREAKING: Anthropic cuts off Claude subscription usage for OpenClaw and third-party AI agents starting April 4, 2026 at 12pm PT. Users must switch to API billing.

VentureBeat2026-04-04

Regressions in opus 4.6 extended - thinking less often even with 'always on'. E.g. acts as if thinking wasn't enabled, even for somewhat difficult tasks where it absolutely should think before answering.

r/ClaudeCode2026-04-01

Nicolas Carlini (67.2k citations on Google Scholar) says Claude is a better security researcher than him. Opus 4.6 found 500+ previously-unknown zero days in open source code.

r/ArtificialInteligence2026-04-04

Core Specs

1000K
Context Window
128K
Max Output
ReasoningOpen Sourcetextimage

Pros & Cons

Sentiment80% +13% ·7% −

Pros

  • +Highest SWE-bench score (80.8%)
  • +128K max output (doubled from 4.5)
  • +Adaptive thinking with effort levels
  • +Agent Teams for parallel coding
  • +Best instruction following in complex contexts

Cons

  • 2x price of GPT-5.4
  • Response prefilling removed (breaking change)
  • Extended thinking deprecated
  • Rate limits can be hit quickly on entry-level plans

Pricing

Input (per 1M tokens)$5.00
Output (per 1M tokens)$25.00
Subscription$20/month
Updated on 2026-04-21

Get Started with Claude

2Choose Pro ($20/mo) for Opus access
3Start chatting or try Claude Code

💡 Free tier uses Sonnet. Upgrade to Pro for Opus.

Benchmarks

sweBenchVerified80.8%
terminalBench265.4%
browseComp84%
gdpvalElo1606%
arcAgi268.8%
gpqaDiamond91.3%
bigLawBench90.2%

Reliability

SLA99.9%
Incidents (30d)45
Last Incident2026-05-08
4/8 故障已解决(监控中)。4月前9天累计 9 次故障,连续3天出问题(4/6-4/8)。
View Status Page →