Claude Opus 4.7 Released — Software Engineering Gains, High-Resolution Vision, Task Budgets [April 2026]

Anthropic made Claude Opus 4.7 generally available on April 16, 2026. Per Anthropic, Opus 4.7 brings notable improvements over Opus 4.6 on difficult software-engineering tasks, substantially better vision support, and higher-quality output for professional artifacts (interfaces, slides, docs).

Headline changes: - Difficult software-engineering tasks: notable gains vs. Opus 4.6 - High-resolution vision: up to 2576px / 3.75MP (a first for Claude) - Task Budgets: a rough token-budget signal for full agentic loops - Higher-quality output for interfaces, slides, and documents - Available in the API, across all Claude products, AWS Bedrock, Google Cloud Vertex AI, and Microsoft Foundry

Pricing is unchanged from Opus 4.6: $5 / $25 per million input/output tokens (as of release). Same-day availability across AWS Bedrock, Google Cloud Vertex AI, and Microsoft Foundry simplifies adoption for enterprise stacks.

Anthropic notes that Opus 4.7's cyber capabilities trail their unreleased internal Mythos Preview, and that the model ships with safeguards to detect and block prohibited or high-risk cybersecurity uses. For high-risk domains, verify Anthropic's official usage policy and allowed use cases before adoption.

Oflight uses Claude Code as the standard pair-programmer in our DocDD-driven AI development workflow. Opus 4.7's gains on hard software tasks feed straight into day-to-day delivery quality. Opus 4.7 is also a cloud backend option for OpenClaw, reserved for tasks where peak quality matters (design judgment, code review, high-quality document generation).

Q1: Should we move from Opus 4.6 to 4.7? A: Pricing is unchanged and difficult-task quality improves — worth evaluating. Re-test your existing prompts for regressions before flipping production. Q2: What's the high-resolution vision for? A: 2576px / 3.75MP enables precise analysis of CAD-style drawings, slide screenshots, scanned documents, and detailed UI mocks. Q3: Why care about Task Budgets? A: It gives Claude a rough sense of total tokens to spend across an agentic loop — useful for cost forecasting and runaway prevention.

- Introducing Claude Opus 4.7 (Anthropic) - Claude Opus 4.7 is generally available (GitHub Changelog, 2026/4/16) - What's new in Claude Opus 4.7 (Claude API docs) - Claude Opus 4.7 in Amazon Bedrock (AWS)