Anthropic's Powerful Opus 4.1 Model Is Here - How To Access It (and Why You'll Want To)

2 hours ago

ZDNET's cardinal takeaways

Anthropic launched Claude Opus 4.1.
The exemplary exceeds nan predecessor's capacity connected analyzable tasks.
It is disposable to paid Claude users, Claude Code, API, Amazon Bedrock, and Google Cloud's Vertex AI.

In May, Anthropic released Claude Opus 4, which nan institution dubbed its astir powerful exemplary yet and nan champion coding exemplary successful nan world. Only 3 months later, Anthropic is upping nan ante further by launching nan highly anticipated Claude Opus 4.1, which now takes its predecessor's crown arsenic Anthropic's astir precocious model.

The Opus family of models is nan company's astir advanced, intelligent AI models geared toward tackling analyzable problems. As a result, Claude Opus 4.1, released connected Tuesday, excels astatine those tasks and tin moreover one-up its predecessor connected agentic tasks, real-world coding, and reasoning, according to Anthropic.

The exemplary besides comes arsenic nan manufacture is expecting nan motorboat of OpenAI's GPT-5 soon.

Also: OpenAI could motorboat GPT-5 immoderate infinitesimal now - what to expect

How does Claude Opus 4.1 perform?

One of nan astir awesome usage cases of Claude Opus 4 was its capacity connected nan SWE-bench Verified, a human-filtered subset of nan SWE-bench, a benchmark that evaluates LLMs' abilities to lick real-world package engineering tasks originated from GitHub. Claude Opus 4's capacity connected nan SWE-bench Verified supported nan declare that it was nan "best coding exemplary successful nan world." As seen successful nan station above, Opus 4.1 performed moreover higher.

Claude Opus 4.1 besides swept its preceding models crossed nan benchmark board, including nan MMMLU, which tests for multilingual capabilities; AIME 2025, which tests for rigor connected precocious schoolhouse lucifer title questions; GPQA, which tests for capacity connected graduate-level reasoning prompts; and more. When pinned against competitors' reasoning models, including OpenAI o3 and Gemini 2.5 Pro, it outperforms them successful various benchmarks, including SWE-bench Verified.

With nan release, Anthropic besides posted its system card, which delineates each of nan information assessments and evaluations it conducted connected nan model, arsenic good arsenic its weaknesses, risks, and limitations. A speedy overview of nan 22-page archive shows that nan exemplary was deployed pinch an AI Safety Level 3 (ASL-3) Standard nether Anthropic's Responsible Scaling Policy (RSP), and it is still prone to astir of nan aforesaid vulnerabilities.

How tin you entree Claude Opus 4.1?

If you want to effort nan exemplary for yourself, it is now disposable to everyone via nan paid Claude plans, which see Claude Pro for $20 per period and Claude Max for $100 per month. It is disposable successful Claude Code, nan API, Amazon Bedrock, and Google Cloud's Vertex AI.

Get nan morning's apical stories successful your inbox each time pinch our Tech Today newsletter.

Editorial standards