News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
imgClaude Opus 4.7

Claude Opus 4.7

Anthropic's large model for long-running coding and professional work

Claude Opus 4.7 is a general-purpose language model from Anthropic, released in November 2026 as the successor to Opus 4.6. The company positions it as its strongest model for advanced software engineering, with improved vision and better handling of long, complex tasks.

Share on WhatsApp

  • Released November 2026 as the successor to Claude Opus 4.6
  • Priced at $5 per million input tokens, $25 per million output tokens
  • First model to trial Anthropic's new cybersecurity safeguards
  • Reads images at higher resolution than Opus 4.6
  • Available via Claude apps, API, Bedrock, Vertex AI and Microsoft Foundry

What it is

Claude Opus 4.7 is a proprietary language model that Anthropic made generally available in November 2026. It sits in the company's Claude line and follows Opus 4.6. Anthropic describes it as aimed at demanding software engineering, along with professional tasks such as building interfaces, slides and documents. The company says it handles long, involved jobs with steady rigour, follows instructions closely and checks its own work before answering. Image handling is sharper than in the previous version because pictures are read at higher resolution. Anthropic also notes it is less broadly capable than Claude Mythos Preview, its most powerful model.

How it came about

Anthropic announced Opus 4.7 as its newest model and made it available the same day across Claude products, its API, Amazon Bedrock, Google Cloud's Vertex AI and Microsoft Foundry. The release followed a week after the company's Project Glasswing statement about the risks and benefits of AI in cybersecurity. There Anthropic said it would keep Claude Mythos Preview on limited release and trial new cyber defences on weaker models first. Opus 4.7 is the first model used for that trial. Early-access testers were quoted at launch, including staff from fintech, legal, developer-tool and agent companies.

Inside the model

Anthropic gives no parameter count, training data, context window or architecture detail for Opus 4.7. What the company does say is that the model was tuned for sustained reasoning over long runs, precise instruction following and self-verification of outputs before reporting. Vision input is processed at higher resolution than in Opus 4.6. During training the team tried ways to reduce cybersecurity capability specifically, and the shipped model carries safeguards that detect and block requests pointing to prohibited or high-risk cyber use. Anthropic says lessons from this deployment will inform a later broad release of Mythos-class models.

What it can do

Anthropic reports gains over Opus 4.6 on advanced software engineering, especially the hardest tasks, and across a range of benchmarks, though it remains below Claude Mythos Preview in general capability. Testimonials cited by the company describe better planning, tool use, long-context consistency and error recovery. Reported figures include 90.9% on BigLaw Bench at high effort for Harvey, over 70% on CursorBench versus 58% for Opus 4.6, a 13% lift on a 93-task coding set, and three times more production tasks resolved on Rakuten-SWE-Bench. One tester said low-effort Opus 4.7 roughly matches medium-effort Opus 4.6.

How to use it

The model is sold through the Claude apps and API, and hosted on Amazon Bedrock, Google Cloud's Vertex AI and Microsoft Foundry. Developers call it as claude-opus-4-7. Pricing is unchanged from Opus 4.6 at $5 per million input tokens and $25 per million output tokens. Weights are not downloadable; access is proprietary. Security professionals doing legitimate work such as vulnerability research, penetration testing and red-teaming can apply to a Cyber Verification Program.

Safety and testing

Anthropic framed this release around cybersecurity risk. It says Opus 4.7's cyber abilities are weaker than those of Claude Mythos Preview, and that training included experiments to reduce them further. The model ships with automated safeguards that detect and block requests indicating prohibited or high-risk cybersecurity use. The company says it wants to learn from real-world use of these controls before widening access to Mythos-class models. A Cyber Verification Program lets vetted security practitioners use the model for permitted work.

Reception and debate

Early-access users quoted by Anthropic praised the model's coding depth, autonomy, instruction following and design judgement. Reported improvements include fewer tool errors, better recall in code review, resistance to getting stuck in loops, and coherent work over several hours in an agent setting. One tester said it pushes back in technical discussions rather than simply agreeing. These are company-selected testimonials rather than independent reviews, and Anthropic published no comparative ranking of its own.

Why it matters

Anthropic presents Opus 4.7 as a step in a staged release strategy: a less capable model carries new cybersecurity safeguards first, and what is learned feeds into a later broad release of Claude Mythos Preview. The company also positions it as a model developers can hand difficult coding work to without close supervision. Testimonials describe it being used for code review, agent orchestration, patent workflows, legal document work, dashboards and long-running investigation tasks.

Questions people ask about Claude Opus 4.7

What is Claude Opus 4.7 for?

Anthropic built it for demanding software engineering and professional work such as interfaces, slides and documents. The company says it suits long, complex jobs because it follows instructions closely, checks its own output and keeps going over extended runs without close supervision.

How is it different from Opus 4.6?

Anthropic reports the biggest gains on the hardest engineering tasks, plus sharper image handling at higher resolution and better judgement on professional work. Testimonials also cite fewer tool errors and stronger long-context consistency. Pricing is unchanged from the earlier version.

How do I get access?

It runs in Claude products and through the Claude API as claude-opus-4-7, and is hosted on Amazon Bedrock, Google Cloud's Vertex AI and Microsoft Foundry. The company charges $5 per million input tokens and $25 per million output tokens. Weights are not released.

Why does Anthropic talk about cybersecurity here?

The company said it would test new cyber defences on weaker models before widening access to Claude Mythos Preview. Opus 4.7 is that test case: training included attempts to reduce cyber capability, and shipped safeguards block requests pointing to prohibited or high-risk cyber use.

Is it Anthropic's most powerful model?

No. Anthropic says Claude Mythos Preview remains its most broadly capable model, and that Opus 4.7 trails it there while beating Opus 4.6 across a range of benchmarks.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Claude Opus 4.7 yet, so the facts on this page come from the sources listed above: anthropic.com/news/claude-opus-4-7. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.