Anthropic Releases Claude Opus 4.7
Anthropic released Claude Opus 4.7 on Thursday, April 16. It's a same-tier upgrade to Opus 4.6, not a new model family, and it comes in at the same list price: $5 per million input tokens and $25 per million output tokens.
The improvements are focused on two areas. The first is long-horizon agentic work, meaning complex, multi-step tasks that run autonomously over an extended period without losing coherence. The second is vision: image resolution capability jumped from 1.15 megapixels to 3.75 megapixels, which translates to meaningfully better performance on image-heavy tasks. Anthropic also added a self-verification loop to how the model works, a plan-execute-verify-report sequence designed to catch errors before surfacing a final output.
On benchmarks, Opus 4.7 ranks second in both coding and reasoning on the major leaderboards, and hit 87.6% on SWE-bench Verified in vendor testing, which is the standard measure for autonomous software engineering tasks.
One thing worth knowing before you update any existing workflows: Anthropic quietly shipped a new tokenizer alongside the model. On code-heavy prompts, the same text may now cost up to 35% more to process, despite the price per million tokens staying flat. If you're running production workflows with significant code volume, it's worth testing actual costs before switching over.
Anthropic also noted that Opus 4.7 is not as capable as Claude Mythos, the model they announced earlier this month and then restricted from public release due to its ability to find security vulnerabilities at a scale that raised serious safety concerns. Opus 4.7 is their most powerful generally available model.
Anthropic Launches Claude Design
Also today, Anthropic launched Claude Design in research preview. It's available now to Pro, Max, Team, and Enterprise subscribers.
Claude Design is a visual creation tool built directly into the Claude interface. You describe what you want, and Claude builds it: prototypes, wireframes, slide decks, one-pagers, social assets, marketing collateral. From there, you can refine through conversation, direct edits, or custom controls. It can pull brand assets from your design files or codebase, and includes a web capture tool that lets it pull colors, typography, and components directly from a live website.
This is not an image generator. It builds structured visual layouts from prompts, closer in spirit to Figma than to Midjourney.
The Canva integration is worth noting. Anything created in Claude Design can be exported as a PDF, PPTX, or URL, or sent directly into Canva where it becomes fully editable and collaborative. Canva, Datadog, and Brilliant are listed as early beta partners.
Anthropic's stated target user is founders, product managers, and marketers who have an idea but not a design background. Experienced designers are also in scope, with the pitch being that it lets them explore more concepts faster than the time constraints of traditional design work typically allow.
The broader pattern here is worth paying attention to. In the past few months, Anthropic has shipped a coding agent, a knowledge-work assistant, desktop computer control, office integrations, a browser agent, and now a design tool. Each one extends the platform further up the stack, into categories that purpose-built software companies have owned for years.
Anthropic Passes OpenAI in Revenue
Anthropic is now ahead of OpenAI in annualized revenue. Anthropic hit $30 billion ARR. OpenAI is at approximately $25 billion.
The context that makes this notable: fifteen months ago, Anthropic was at $1 billion in annualized revenue. That's 30x growth in just over a year.
The company is also reported to be in early conversations with Goldman Sachs, JPMorgan, and Morgan Stanley about a potential IPO, with a timeline that could come as early as October 2026. Nothing is confirmed, and IPO timelines shift, but the direction of travel is clear. Anthropic is no longer positioning as a safety-focused research lab with a product on the side. It is a product company with an enterprise motion that is currently outpacing the most well-funded AI company in the world on revenue.
Apple Is Weighing Anthropic or OpenAI to Power Siri
The Los Angeles Times reported this week that Apple is considering replacing its in-house AI and handing Siri over to either Anthropic or OpenAI.
No decision has been announced and the situation may change, but the signal is worth taking seriously. Apple has over a billion active iPhone users. If either Claude or ChatGPT becomes the default voice behind Siri, that represents a distribution channel for AI that is several orders of magnitude larger than anything currently in play. It would also mean that a significant portion of the population's first real interaction with a frontier AI model happens through an interface they already use every day without thinking about it.
The outcome of this, if it happens, will likely shape consumer AI familiarity in ways that go well beyond what any product launch or marketing effort could accomplish on its own.