Claude Opus 4.7 Is Here: Better Coding, Sharper Vision, and a Surprising Cyber Safety Twist

Anthropic just released Claude Opus 4.7, and it’s a solid upgrade over Opus 4.6—especially if you’re doing heavy coding or working with images. The model is now generally available across all Claude products, the API, Amazon Bedrock, Google Cloud’s Vertex AI, and Microsoft Foundry.

Pricing hasn’t budged: still $5 per million input tokens and $25 per million output tokens. That’s reasonable, given what you’re getting.

What’s Actually Better?

The headline improvement is in advanced software engineering. Anthropic says Opus 4.7 handles the kind of coding work that previously required close supervision—the really gnarly stuff where you’d normally babysit the model. Early testers report handing off their hardest tasks with confidence, and the model catches its own logical faults during planning. That’s a big deal for anyone running complex, multi-step workflows.

Vision also got a real boost. The model can now see images at higher resolution, and it’s noticeably better at interpreting technical diagrams, chemical structures, and other dense visuals. One tester from Solve Intelligence mentioned it’s helping them build tools for life sciences patent workflows—drafting, prosecution, infringement detection, the whole pipeline.

On the creative side, Anthropic claims Opus 4.7 is more tasteful and produces higher-quality interfaces, slides, and docs. I’d take that with a grain of salt until I see it myself, but the trend is encouraging.

The Cyber Safety Angle

Here’s where it gets interesting. Last week, Anthropic announced Project Glasswing, which highlighted both the risks and benefits of AI for cybersecurity. They said they’d keep Claude Mythos Preview’s release limited and test new cyber safeguards on less capable models first. Opus 4.7 is that test case.

Anthropic explicitly states that Opus 4.7’s cyber capabilities are not as advanced as Mythos Preview’s. In fact, during training they experimented with differentially reducing those capabilities. The model ships with safeguards that automatically detect and block requests indicating prohibited or high-risk cybersecurity uses.

If you’re a security professional who needs Opus 4.7 for legitimate work—vulnerability research, penetration testing, red-teaming—you can apply for Anthropic’s new Cyber Verification Program. It’s a bit of a gate, but I get why they’re being cautious.

What Early Testers Are Saying

The feedback is overwhelmingly positive, and the specific numbers are worth noting:

• One tester reported a 13% lift in resolution on a 93-task coding benchmark, including four tasks that neither Opus 4.6 nor Sonnet 4.6 could solve.
• On a research-agent benchmark, Opus 4.7 scored 0.715 overall, tying for the top spot across six modules. Its General Finance score jumped from 0.767 to 0.813.
• Multiple testers highlighted that the model handles long-running, async workflows better than previous versions. One said it “works coherently for hours” and “pushes through hard problems rather than giving up.”
• A tester from Hex noted that Opus 4.7 correctly reports when data is missing instead of making up plausible-but-incorrect fallbacks. That alone is a huge quality-of-life improvement.

The Catch

Opus 4.7 is not Anthropic’s most capable model. That title belongs to Claude Mythos Preview, which was announced last week and is being held back for safety reasons. Opus 4.7 is a step down in raw cyber capability, and Anthropic is upfront about that.

But honestly? For most developers and teams, Opus 4.7 is probably the model you actually want. It’s available now, it’s priced competitively, and it ships with safeguards that make it safer to deploy in real-world environments. Mythos Preview might be the flashy new thing, but Opus 4.7 is the workhorse that’s ready today.

If you’re already using Opus 4.6, the upgrade is a no-brainer. Same price, better performance, and you get the new vision capabilities. I’d suggest running your own benchmarks on the tasks that matter to you, but the early data looks strong.

Available now on all major platforms. Go test it.