Anthropic shipped Claude Opus 4.7 today, April 16, 2026. No fanfare event. No keynote. Just a point release that quietly moved the ceiling up again.
If you’re already paying for Claude, you don’t need to do anything — you’re already using it. If you’re not, here’s why that changes the math again.
What Actually Changed
Five things worth knowing, in plain English:
1. 13% better at coding than Opus 4.6.
Anthropic’s own pitch: you can “hand off your hardest coding work.” If you’re a business running custom scripts, Zapier-style automations, or anything your internal tech person wrestles with — that work just got measurably easier.
2. State-of-the-art on the “office work” benchmark.
GDPval-AA is the test that measures finance, legal, and general knowledge-work tasks. Opus 4.7 posts the best public scores anyone has on it. For small businesses, that’s the relevant benchmark. Not physics PhDs. Bookkeeping. Contracts. Research. Summaries.
3. Vision got 3x better.
It can now read images up to 2,576 pixels on the long edge — over 3x what previous Claude models could handle. Scanned receipts, handwritten notes on whiteboards, blueprints, messy PDFs — they work now. If “I tried AI on my scanned documents and it was useless” was your experience in 2025, try again today.
4. It checks its own work.
This is the one people will feel without understanding why. Anthropic built in self-verification — the model “devises ways to verify its own outputs before reporting back.” In practice: fewer confident wrong answers on long tasks. Fewer “AI hallucinations” in your client emails.
5. Same price.
Claude Pro is still $20/month. API pricing is unchanged at $5 per million input tokens, $25 per million output. You get the capability jump without a price jump. Pay attention to this pattern — it keeps happening.
What This Means For Your Business
If you’re on Claude Pro ($20/month): You’re already using Opus 4.7. Go run whatever task you’ve been using Claude for — the output quality you got yesterday is the floor today.
If you use Claude Code or Cowork on your desktop: Update the app. New features shipped today include a /ultrareview command in Claude Code and extended auto mode for Max subscribers. Your AI teammate just got a sharper editor and a longer attention span.
If you’re still on the sidelines: The cost of waiting went up again.
I keep making this point in the Cheat Codes series because it keeps being true. Every month Anthropic, OpenAI, and Google ship upgrades. Every month, the businesses already using these tools pull further ahead on output. The learning curve to catch up doesn’t get shorter — but the gap does get wider.
In January I called the $20 Claude Pro tier “absurdly underpriced for what you get.” In April, at the same price, the tier includes:
- A model that outperforms last year’s most expensive offerings
- Vision that reads messy real-world documents
- Self-verification that reduces hallucinated answers on business work
- Desktop integration that operates alongside you, not just in a chat window
That’s a colleague for $20 a month. I don’t know how else to frame it.
The Honest Part
Anthropic also released something eight days ago called Claude Mythos Preview — a much more powerful model that they will not make available to the public. They gave it to 40+ critical infrastructure partners and kept the rest of us on Opus-tier.
Today’s release is the public version. It’s extremely capable. It is deliberately not the ceiling.
Why does that matter to a Waseca accounting firm or a Montgomery nonprofit? Because it tells you something about the trajectory. The public tier keeps getting better — today’s release is proof. But there’s now a visibly better tier sitting behind a locked door, which means Anthropic’s internal tools are already doing things your public tools can’t.
Translation: the gap between what people building the AI can do and what the rest of us can do is widening. The correct response is to use the public tools hard enough that when the next tier opens up, you’re ready.
What To Do This Week
-
If you haven’t tried Claude in 30+ days: Try it again today. Give it a real task — summarize a contract, draft a client reply you’ve been avoiding, reconcile a messy spreadsheet. The April 16 version is not the version you remember.
-
If you’re already using it casually: Go set up a Claude Project with your business context — your style guide, your recurring document templates, your client list. Context-aware AI is a different category from chat-window AI.
-
If you’re running Claude Code or Cowork: Update your desktop app today. Check the release notes. The small features (
/ultrareview, extended auto mode, Task Budgets in beta) compound fast in daily workflows. -
If you need help setting any of this up: That’s literally what we do. Small-business and nonprofit IT, AI-assisted, Minnesota-based.
The cheat codes are still free. The tools are still $20/month. The game just got faster again.
Stop reading about AI. Start using it.
Matt Stoltz is the founder of Flower Insider Technologies, an AI-assisted managed IT company serving small businesses and nonprofits in southern Minnesota.
Questions? Contact FIT or find me on LinkedIn.
Source: Anthropic’s Claude Opus 4.7 announcement, April 16, 2026 (anthropic.com/news/claude-opus-4-7).