Anthropic has released Claude Opus 4.7, a new AI model designed for advanced coding, high-resolution vision, and long-running agent workflows, now available across Claude, APIs, and major cloud platforms.

The update delivers measurable gains in software engineering, visual reasoning, and multi-step task performance, while introducing built-in cybersecurity safeguards tied to Anthropic's broader safety research initiative.

This matters because AI vendors and enterprise teams are increasingly focused on whether models can reliably execute complex, multi-step workflows with less supervision, and whether those systems can be deployed safely at scale.

The release also adds platform-level controls, including a new xhigh effort setting for reasoning depth, task budgets in public beta, and expanded Claude Code features, reinforcing Anthropic's push toward production-ready AI systems.

The update directly impacts enterprise developers, AI engineering teams, and product leaders building on Claude, particularly those evaluating reliability, cost control, and safety in real-world deployments.

In short, Claude Opus 4.7 is a production-focused upgrade that improves coding autonomy, visual precision, and task reliability while introducing the first real-world deployment of Anthropic's cybersecurity safeguard architecture.

Claude Opus 4.7 is Anthropic's latest frontier AI model designed for complex, long-running tasks, combining advanced reasoning, high-resolution vision, and built-in safeguards for controlled deployment.

Claude Opus 4.7 is Anthropic's latest production AI model, designed to improve coding, vision, and agent reliability while introducing built-in cybersecurity safeguards for enterprise deployment.

Anthropic announced Claude Opus 4.7 as now generally available, describing it as a direct and substantial upgrade to Opus 4.6 in the Claude 4 model family. According to Anthropic, the model is engineered to handle complex, long-running tasks with greater rigor and consistency, pays closer attention to instructions than its predecessor, and is designed to verify its own outputs before reporting results — a behavior the company says allows users to hand off the hardest coding work that previously needed close supervision.

The model is also more creative and tasteful when completing professional tasks, producing higher-quality interfaces, slides, and docs, and brings substantially better vision with support for images at greater resolution than prior Claude models.

The model is available immediately across all Claude consumer products, the Claude API, Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry, making it accessible to both individual developers and enterprise teams across major cloud infrastructure providers. Pricing is unchanged from Opus 4.6: $5 per million input tokens and $25 per million output tokens. Developers can access the model via the API using the model string claude-opus-4-7.

While Anthropic describes Opus 4.7 as less broadly capable than its most powerful model, Claude Mythos Preview, it outperforms Opus 4.6 across a wide range of benchmarks:

The benchmark results place Opus 4.7 ahead of Opus 4.6 across every measured category, from coding and reasoning to vision and financial analysis.

Several capability improvements in Opus 4.7 extend beyond raw benchmark performance. The most significant is enhanced high-resolution image support: the model now accepts images up to 2,576 pixels on the long edge, or approximately 3.75 megapixels — more than 3 times the resolution limit of prior Claude models. This is a model-level change rather than an API parameter, meaning images sent to Claude will simply be processed at higher fidelity automatically. Anthropic notes that higher-resolution images consume more tokens, so users who do not require the extra detail can downsample images before sending them.

The practical impact of this vision improvement is significant for use cases that depend on fine visual detail. Computer-use agents reading dense application screenshots, developers performing data extraction from complex diagrams, and workflows that require pixel-perfect visual references can now operate at a fidelity that previous Claude models could not support. XBOW, an autonomous penetration testing company, reported that Opus 4.7 scored 98.5% on their visual-acuity benchmark, compared to 54.5% for Opus 4.6 — a result they described as effectively eliminating their single biggest Claude pain point and unlocking an entire class of computer-use work.

Instruction following is another area of substantial improvement. Anthropic says Opus 4.7 takes instructions more literally than its predecessors, which means prompts written for earlier models can produce unexpected results when run against Opus 4.7 — where previous models might have interpreted instructions loosely or skipped parts, Opus 4.7 executes them precisely. Anthropic recommends that developers re-tune their prompts and harnesses accordingly when migrating.

Opus 4.7 also improves on file system-based memory, retaining important notes across long, multi-session work and using them to reduce the up-front context required on subsequent tasks.

Anthropic's internal testing showed Opus 4.7 performing as a more effective finance analyst than Opus 4.6, producing more rigorous analyses and financial models, more professional presentations, and tighter integration across complex tasks. Opus 4.7 is also state-of-the-art on GDPval-AA, a third-party evaluation of economically valuable knowledge work across finance, legal, and other domains.

Anthropic is embedding cybersecurity safeguards directly into Opus 4.7 as part of Project Glasswing, the company's research and policy initiative examining both the risks and benefits of AI models for cybersecurity. As part of that initiative, Anthropic committed to keeping Claude Mythos Preview in limited release while testing new safeguards on less capable models first.

Opus 4.7 is the first model to carry that architecture into production, making this release a defining moment in Anthropic's approach to responsible cybersecurity deployment.

During training, Anthropic experimented with efforts to differentially reduce the model's cyber capabilities relative to Mythos Preview. The production release includes automated safeguards that detect and block requests that indicate prohibited or high-risk cybersecurity uses. Anthropic says the data it collects from real-world deployment of these safeguards will inform its path toward eventually making Mythos-class models more broadly available.

For security professionals with legitimate use cases — including vulnerability research, penetration testing, and red-teaming — Anthropic is launching a Cyber Verification Program through which professionals can apply for verified access to Opus 4.7's cybersecurity capabilities.

Early-access feedback from enterprise partners provides concrete, real-world performance data across a range of professional environments — with each reporting specific, measurable improvements on their own production benchmarks.

Intuit VP of Technology Clarence Huang said the results from early testing point to a meaningful leap for developers. “In early testing, we’re seeing the potential for a significant leap for our developers with Claude Opus 4.7. It catches its own logical faults during the planning phase and accelerates execution, far beyond previous Claude models. As a financial technology platform serving millions of consumers and businesses at significant scale, this combination of speed and precision could be game-changing: accelerating development velocity for faster delivery of the trusted financial solutions our customers rely on every day.”

GitHub Chief Product Officer Mario Rodriguez reported that on the company's 93-task coding benchmark, Opus 4.7 delivered a 13% lift in resolution over Opus 4.6 — including 4 tasks that neither Opus 4.6 nor Sonnet 4.6 could solve. “Combined with faster median latency and strict instruction following, it's particularly meaningful for complex, long-running coding workflows. It cuts the friction from those multi-step tasks so developers can stay in the flow and focus on building.”

Notion AI Lead Sarah Sachs reported that for complex multi-step workflows, Opus 4.7 delivered 14% improvement over Opus 4.6 at fewer tokens and with one-third the tool errors. “It’s the first model to pass our implicit-need tests” — tests that measure whether a model can infer what a user needs without it being explicitly stated — “and it keeps executing through tool failures that used to stop Opus cold. This is the reliability jump that makes Notion Agent feel like a true teammate.”

Ramp Software Engineer Austin Ray highlighted the model's performance in agent-team environments. “For Ramp, Claude Opus 4.7 stands out in agent-team workflows. We're seeing stronger role fidelity, instruction-following, coordination, and complex reasoning, especially on engineering tasks that span tools, codebases, and debugging context. Compared with Opus 4.6, it needs much less step-by-step guidance, helping us scale the internal agent workflows our engineering teams run.”

The Opus 4.7 release is accompanied by a set of platform and product updates that extend across the Claude API, Claude Code, and Claude consumer products.

Opus 4.7 introduces a new xhigh (extra-high) effort level — an intermediate setting between high and max in Anthropic's effort control system. This gives developers finer control over the tradeoff between reasoning depth and latency on hard problems. Anthropic has raised the default effort level in Claude Code to xhigh for all plans, and recommends developers start at high or xhigh when testing Opus 4.7 for coding and agentic use cases.

On the Claude Platform API, Anthropic is also launching task budgets in public beta, a mechanism that gives developers a way to guide Claude's token spend so it can prioritize work across longer task runs.

In Claude Code, Anthropic is introducing the /ultrareview slash command, which produces a dedicated review session that reads through code changes and flags bugs and design issues that a careful reviewer would catch. Pro and Max Claude Code users receive 3 free ultrareviews at launch. Anthropic is also extending auto mode to Max users — a permissions option that allows Claude to make decisions on a user's behalf during long-running tasks, reducing the number of interruptions required without requiring the user to skip all permissions manually.

Developers migrating from Opus 4.6 to Opus 4.7 should plan for 2 specific changes that affect token usage. First, Opus 4.7 uses an updated tokenizer designed to improve how the model processes text. The practical tradeoff is that the same input can map to more tokens — roughly 1.0–1.35× depending on content type. Second, Opus 4.7 thinks more at higher effort levels, particularly on later turns in agentic settings, which improves reliability on hard problems but also increases output token counts.

Anthropic offers several mechanisms for managing token usage: the effort parameter, task budgets, and prompting the model to be more concise. The company's own testing on an internal coding evaluation shows net-favorable results across all effort levels, but recommends that developers measure the difference on real traffic before fully deploying. A dedicated migration guide is available through the Claude Platform documentation.

Q: What is Claude Opus 4.7 and what was announced?
A: Claude Opus 4.7 is Anthropic's latest production AI model, released as a direct upgrade to Opus 4.6. It delivers improved performance across coding, high-resolution vision, instruction following, and agentic workflows, while introducing built-in cybersecurity safeguards and new platform-level controls.

Q: How does Claude Opus 4.7 improve performance and capabilities?
A: The key point: Opus 4.7 improves reliability on complex, multi-step tasks by combining stronger instruction following, higher-resolution vision, and better long-running task execution. It achieves higher benchmark scores across software engineering, visual reasoning, and agentic workflows, while enabling more precise execution of detailed instructions.

Q: Why does Claude Opus 4.7 include cybersecurity safeguards?
A: The safeguards are part of Anthropic's Project Glasswing initiative, which focuses on managing AI cybersecurity risks. Opus 4.7 is the first production model to include automated detection and blocking of high-risk or prohibited cybersecurity use cases, while allowing verified professionals access through a controlled program.

Q: What should developers expect when migrating to Opus 4.7?
A: Developers may see increased token usage due to a new tokenizer and higher reasoning effort levels. Anthropic recommends using effort controls, task budgets, and testing on real workloads to manage costs and performance tradeoffs.

Q: What additional features are launching alongside Opus 4.7?
A: Anthropic introduced a new xhigh effort level, task budgets in public beta, the /ultrareview feature in Claude Code, and expanded auto mode capabilities, all aimed at improving control, efficiency, and long-running task execution.

The release of Claude Opus 4.7 combines measurable performance gains with a structured cybersecurity safeguard architecture designed for real-world enterprise deployment and testing.

Key point: Claude Opus 4.7 combines measurable performance gains with the first production deployment of Anthropic's cybersecurity safeguard architecture, moving AI systems closer to reliable, real-world execution of complex tasks.

Who should care: Enterprise developers, AI engineering teams, and product leaders building with Claude should evaluate this release closely. The reported gains from partners including GitHub, Notion, Cursor, Ramp, and Devin reflect measurable improvements on real production workflows.

Why this matters now: The gap between AI systems that require constant supervision and those that can reliably complete complex, multi-step work is becoming a defining competitive factor. Opus 4.7 directly targets that gap through improved instruction fidelity, reduced tool errors, and stronger long-horizon task performance.

What decision this affects: Teams deciding whether to expand AI agent usage or trust models with unsupervised workflows now have clearer evidence to reassess prior limitations. Opus 4.7 provides measurable evidence that reliability thresholds are improving in production environments.

In short, Claude Opus 4.7 strengthens the case for deploying AI in high-stakes, multi-step workflows by improving reliability, visual precision, and safety controls — while giving enterprises clearer pathways to scale AI responsibly.

The next question is not whether Opus 4.7 is better — the data says it is — but whether the industry is ready to hand AI the keys to its most complex workflows.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing support from Claude, and AEO/GEO/SEO optimization, image concept development, and editorial structuring support from ChatGPT, AI assistants. All final editorial decisions, perspectives, and publishing choices were made by Alicia Shapiro.