Review

Claude: Strong on writing, less settled than it looks

Claude is among the strongest AI assistants for writing, reasoning, and coding, but its most useful features now sit behind a pricing and privacy structure that demands attention.

Last updated April 2026 · Pricing and features verified against official documentation

Claude is the assistant people reach for when they want something more measured than flashy. Its reputation came from clean long-form prose, keeping a thread across sprawling documents, and staying composed when the task turns analytical.

Anthropic has also turned Claude into a broader product than it was at launch. What started as a chat interface now includes Claude Code, Research, Projects, Workspace connections, remote MCP support, and agent features for office work. That leaves Claude layered.

That expansion is mostly a good thing. If you want an assistant that can draft, revise, reason across long context, and handle serious coding tasks without drifting into generic AI mush, Claude is one of the strongest options available. Anthropic’s newest generally available flagship is now Opus 4.7; the company has also described Mythos Preview as more powerful, but limited enough that it should not be treated as the normal Claude buying story.

The catch is that Claude is increasingly sold in layers. The most interesting coding and agent features are not evenly distributed, and consumer privacy defaults require more attention than many users expect. Claude is excellent, but it is no longer simple.

What the Product Actually Is Now

Claude is most usefully understood as a family of products wrapped in one interface. At the top sits Claude itself, now spread across web, desktop, iOS, and Android, with a free tier that includes more than older Claude plans did: web search, memory, code execution, file creation, and connectors. Beneath that sits a paid ladder that unlocks more usage, more models, and Anthropic’s highest-end work surfaces.

Around that core Anthropic has built Projects, Research, Google Workspace and Slack connections, remote MCP support, Claude Cowork, and Claude Code. Opus 4.7 is now the premium generally available model Anthropic surfaces for Pro, Max, Team, and Enterprise users, which matters because Claude increasingly serves two buyers at once: one wants a thoughtful assistant for writing and research, and the other wants an agentic system that can plan, code, and operate across tools with less supervision.

Strengths

Polished long-form writing. Claude still sets the standard for clean first drafts that sound like they were written by a human who cares about rhythm, not just correctness. It is especially strong on memos, analysis, reports, and editorial work where tone matters as much as content. ChatGPT can match it on breadth, but Claude usually wins on prose quality and restraint.

Long-context reasoning that actually feels useful. Opus 4.7 is now marketed with a 1M context window, and that extra room is not just a benchmark badge. It makes Claude meaningfully better at large document sets, long research threads, and codebases that would otherwise need to be chopped into pieces. Mythos Preview may be the more powerful research model, but Opus 4.7 is the relevant model for most buyers because it is the broadly available one.

Claude Code is a real reason to buy in. Claude is unusually good at planning, refactoring, and staying coherent across a project, especially when you want the model to work inside an existing codebase rather than generate toy examples. Anthropic now treats Claude Code as a first-class surface across terminal, desktop, IDE, and team plans, which makes Claude feel less like a chatbot with coding bolted on and more like a serious development product.

Power without the platform sprawl. Claude does not try to be an all-singing platform in the way ChatGPT does, and that restraint helps. Projects, Research, web search, and integrations stay useful without turning the interface into a control panel, which is a real advantage for users who want power without feature sprawl.

Weaknesses

The most useful features are unevenly distributed. Claude’s coding and agent features are not evenly available across tiers, and that makes the product harder to evaluate than it should be. Free now does more real work than it used to, but Opus 4.7, Claude Code, Claude Cowork, and the higher-usage seats are where Claude starts to look like a platform for power users and procurement teams rather than a simple assistant.

Consumer privacy defaults are not forgiving. Anthropic makes consumer users explicitly choose how chats and coding sessions can be used to improve Claude. That is better than burying the issue, but it also means the free, Pro, and Max experience is not the same as the commercial one. If you are handling client work, internal research, or anything confidential, the consumer tiers are the wrong place to assume they are enough.

The ecosystem is still smaller than the giants’. Claude’s integrations are useful, but they do not match the breadth of Microsoft’s or Google’s surrounding products. Gemini has the tighter pull into Workspace, Microsoft Copilot has the obvious Microsoft gravity, and Claude’s network effects are weaker even when its model quality is stronger.

Pricing

Claude’s pricing is easiest to read as a ladder from broad free assistant to expensive power-user service. Free now covers enough ground to test Claude seriously. Pro remains the obvious individual buy at $20 per month or $200 per year, because it adds more usage, Claude Code, Claude Cowork, Research, unlimited projects, and access to more models without forcing a business purchase.

Max is for people who keep hitting Pro limits and are willing to pay to avoid friction. Team now starts at $20 per seat per month on annual billing for Standard seats and $100 for Premium seats, which is more legible than the older split but still expensive once you need premium access across a wider group. Enterprise is now framed even more directly as $20 per seat plus usage billed at API rates.

That structure makes Claude look less like a universal subscription and more like a ladder for users whose tolerance for rate limits has already been tested. If Claude is central to your workday, the ladder is understandable. If it is not, the premium seats and usage-based enterprise layer get expensive quickly.

Privacy

This is the section where Claude asks for real attention. On consumer plans, Anthropic says Free, Pro, and Max users choose whether chats and coding sessions can be used to improve Claude, including when Claude Code is used under those accounts. Anthropic also says flagged conversations may still be used for safety review, and that raw content from connectors is not used for model improvement unless it is copied directly into the conversation.

The practical takeaway is simple: if your work involves sensitive documents, proprietary information, or regulated data, the commercial tiers are the better default. Anthropic says Team, Enterprise, and the API do not train on customer data by default, and its commercial documentation lists SOC 2 Type I and Type II, ISO 27001:2022, ISO/IEC 42001:2023, and a HIPAA-ready configuration with a BAA available.

Who It’s Best For

Writers who want strong first drafts and minimal fuss. Claude is excellent for analytical prose, editing, synthesis, and long-form content that needs to sound measured rather than exuberant.

Researchers and analysts working across long documents. The long context window and Project-style organisation make Claude a strong fit for dense source material, recurring client work, and report-heavy workflows.

Developers who want an AI pair programmer that can sustain a task. Claude Code is one of the strongest reasons to choose Anthropic, especially for longer refactors, code review, and multi-file reasoning.

Teams that care about governance but do not want a bloated platform. Team and Enterprise plans give Claude a real business posture without forcing you into the broader product sprawl of Microsoft or Google.

Who Should Look Elsewhere

Users who want the biggest app ecosystem should compare ChatGPT first. It is more sprawling, but that sprawl buys breadth and a larger surrounding ecosystem.

Google-centric teams will often be happier in Gemini, especially if the buying decision is tied to Workspace, Gmail, Docs, and storage bundles.

Research-first users who care most about search and citations should still evaluate Perplexity. Claude can do research well, but it is not as cleanly built around that workflow.

Teams that need the cheapest collaborative AI stack may find Claude’s premium plans too expensive once seat minimums and premium tiers enter the picture.

Bottom Line

Claude makes a strong case for a quiet AI assistant that respects the shape of serious work. Opus 4.7 strengthens that case rather than changing it: Claude still writes better than most rivals, reasons more carefully than many of them, and now has enough coding and agentic capability to justify a real seat at the table.

The tradeoff is that Anthropic has turned a graceful product into a more stratified business. The consumer tiers are strong but privacy-sensitive, the premium tiers are expensive, and the ecosystem is still not as broad as the competition’s.

It is not the most expansive AI product. It is one of the better ones to actually use.