AI Daily 18 Apr: Claude Opus 4.7 and Anthropic’s Compute Bet

A busy week from Anthropic closes with a sharper model, a new design product, and a compute deal that signals the scale of what the company is planning. Meanwhile the UK’s financial regulator presses ahead with its crypto rulebook and sets out a vision for open finance.

Anthropic has released Claude Opus 4.7, a notable step up from Opus 4.6 with particular gains in advanced software engineering and a substantially upgraded ability to read images. The model is available now across claude.ai, the Anthropic API, Amazon Bedrock, Google Cloud’s Vertex AI, and Microsoft Foundry. Pricing remains the same as Opus 4.6: $5 per million input tokens and $25 per million output tokens.

The headline improvement is vision. Opus 4.7 can accept images up to 2,576 pixels on the long edge, roughly 3.75 megapixels and more than three times the resolution supported by previous Claude models. That makes it meaningfully more useful for tasks such as reading dense screenshots, extracting data from complex diagrams, or referencing detailed visual documents. The model is described by Anthropic as “more tasteful and creative” when producing professional outputs like interfaces, slides, and documents.

On the coding side, Anthropic says early users are reporting they can hand off their hardest software engineering work, the kind that previously needed close supervision, and trust the model to complete it with rigour. Opus 4.7 also devises ways to verify its own outputs before reporting back, which matters a great deal in agentic workflows where a long chain of steps can go wrong quietly. The model sits below “Claude Mythos Preview” in Anthropic’s current lineup in terms of raw capability, but is positioned as the best practical choice for most demanding work today.

Alongside the model release, Anthropic Labs has announced Claude Design, a new product aimed at generating high-quality visual and document output. Anthropic Labs is the company’s in-house product studio, distinct from the core API business, and Claude Design appears to be targeted at professionals who need polished interfaces, presentations, and written materials rather than raw code. Few technical details have been published at the time of writing, but the launch fits a pattern of Anthropic moving further into the end-user application space rather than leaving that ground entirely to third-party developers.

The practical question for small businesses and freelancers is whether Claude Design functions as a standalone creative tool or as a feature layer inside Claude.ai. More detail is expected to follow, and it is worth keeping an eye on the Anthropic Labs page for updates over the coming days.

Anthropic has also announced an expansion of its infrastructure partnership with Google and Broadcom, targeting multiple gigawatts of next-generation compute capacity. Gigawatts is a power measurement: a single gigawatt can support tens of thousands of high-performance AI chips running simultaneously. The announcement is a statement of intent about the scale at which Anthropic expects to be operating in the coming years, and it places the company firmly in the same infrastructure arms race as its larger rivals. For users, this kind of compute commitment is what ultimately determines how quickly more capable models can be trained and deployed.

Google has been a significant backer of Anthropic since 2023 and the Broadcom relationship is newer, focusing on custom silicon development. The specific timeline and total investment figure for the expanded deal were not disclosed in the announcement.

The UK’s Financial Conduct Authority has opened a consultation on guidance for the country’s forthcoming crypto asset regulation framework. Published on 15 April, the consultation is an important step in the process of turning the UK’s new crypto laws into practical rules for firms. The FCA is asking for industry feedback on how authorisation, conduct standards, and consumer protection should work under the regime. Firms operating in the UK that hold or handle crypto assets will need to be authorised under the new framework, and this consultation sets out how the FCA expects that to look in practice.

The FCA has positioned the UK as wanting to be a credible home for well-regulated crypto businesses, rather than an open door or a closed one. The consultation period gives firms and consumer groups a chance to shape the detail before rules are finalised. Anyone in the crypto space with UK customers or operations should read the document and consider responding. The FCA’s website has the consultation paper and response form.

Separately, the FCA set out its vision for open finance on 14 April, describing how wider data sharing in financial services could benefit both consumers and small businesses. Open banking, which allows customers to share their bank account data with third-party apps, has been running in the UK for several years. Open finance extends that concept to other financial products, including savings, investments, pensions, and mortgages. The FCA’s vision paper explores how giving consumers better control over their financial data could enable more personalised products and more meaningful price comparison. For small businesses in particular, better access to their own financial data could make it easier to manage cash flow, access credit, and switch providers.

Here is everything else worth knowing from today’s AI news.

• Anthropic has launched The Anthropic Institute, a new body focused on AI safety research and policy engagement, signalling a more structured approach to the company’s public-interest work. Source
• Anthropic has committed $100 million to the Claude Partner Network, a programme supporting businesses that build products and services on top of Claude, covering tools, go-to-market support, and co-selling arrangements. Source
• The Australian government has signed a memorandum of understanding with Anthropic covering AI safety research and responsible deployment, making Australia the latest country to formalise a government-level relationship with a major AI lab. Source
• Anthropic surveyed 81,000 people to understand what users actually want from AI, publishing findings that touch on privacy, accuracy, and the desire for AI that supports rather than replaces human judgement. Source
• The FCA published Year 2 Consumer Duty board report findings, setting out where financial services firms are making progress and where gaps remain in their obligations to retail customers under the rules introduced in 2023. Source
• Meta’s SAM 3.1 (Segment Anything Model 3.1) brings faster real-time video object detection using multiplexing and global reasoning, making the open-source tool more practical for developers building video analysis applications. Source

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s daily guide to developments in artificial intelligence — published every evening.