It’s official: Claude Opus 4.7 is here. After The Information leaked the launch yesterday, Anthropic made the formal announcement today. And the new flagship model delivers.
What Opus 4.7 does better
The biggest leap is in coding. Anthropic reports a 13% improvement on a 93-task coding benchmark and solved four tasks that both Opus 4.6 and Sonnet 4.6 had failed. Enterprise customers report 10-15% higher success rates on engineering workflows.
Those sound like incremental numbers, but in practice it means you can hand Opus 4.7 complex, long-running tasks and trust it to execute consistently. Less micromanagement, less fixing things after the fact — exactly what power users want to hear after the nerfing complaints of recent weeks.
Triple the image capacity
A feature that’s not getting enough attention: Opus 4.7 processes images up to 2,576 pixels on the long edge — roughly 3.75 megapixels. That’s more than triple what previous Claude models could handle. Dense screenshots, complex diagrams, pixel-precise UI reviews — all dramatically more usable.
New effort level: xhigh
Particularly interesting given the recent effort-level debate: Anthropic introduces a new level called ‘xhigh’ that sits between ‘high’ and ‘max’. Developers now have finer control over how much thinking Claude puts into a task. On top of that, Task Budgets are now in public beta, letting you manage token spending across extended sessions.
What else changes
Opus 4.7 also brings Claude Code updates: Ultrareview is a new slash command for code reviews (Pro/Max users), Auto Mode extends to Max users, and File System Memory improves context retention across multi-session work.
One thing Anthropic is upfront about: the updated tokenizer processes text differently and may increase token usage by 1.0-1.35x depending on content type. However, net efficiency on coding tasks should still improve.
Pricing and availability
Prices stay the same: $5 per million input tokens, $25 per million output tokens. Available via the Claude API (claude-opus-4-7), Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.
Security and Glasswing
Given the Mythos discussion, Anthropic has equipped Opus 4.7 with automatic safeguards that detect and block high-risk cybersecurity requests. Security researchers can register for the new Cyber Verification Program for legitimate work like penetration testing.
My take
Opus 4.7 isn’t a revolutionary leap — but it’s exactly the right update at exactly the right time. Anthropic was under pressure: the nerfing allegations, the outages, the pricing debates. This release shows the company is working on model quality, not just marketing.
The 13% coding improvement sounds modest, but anyone who works with Claude Code daily will feel the difference. And the new xhigh level is an elegant admission: yes, we adjusted the effort — here’s the solution.
Sources: