Claude Opus 4.7 Released — Opus 4 Deprecated, Migrate Now

Anthropic shipped Claude Opus 4.7 on April 16, 2026 — the highest-capability model in the Claude lineup. The upgrade brings sharper image analysis and stronger performance on complex, long-running coding tasks that engineering teams and AI automation developers increasingly depend on. At the same time, both Opus 4 and Opus 4.1 have been permanently deprecated (retired and removed from availability), forcing users to migrate immediately.

What Claude Opus 4.7 Actually Changes

The headline improvements cover two areas where developers have pushed hardest: vision and software engineering.

On the vision side — Claude's ability to analyze images, diagrams, and screenshots — Opus 4.7 processes images at higher resolution than any previous Opus release. Anthropic's documentation states: "Opus 4.7 shows improvements in software engineering and complex, long-running coding tasks, as well as better vision, allowing it to see images in higher resolution." In practice, this means the model can read smaller text inside screenshots, identify finer details in architecture diagrams, and process dense data tables without losing accuracy at image edges.

On the coding side, Opus 4.7 is specifically optimized for long-running tasks — multi-step engineering workflows that unfold over hours rather than minutes. This matters for agentic (AI-driven, multi-step automated) use cases: code reviews across large codebases, automated refactoring projects, vibe coding sessions where developers describe intent and Claude generates full implementations, and extended debugging sessions where earlier context needs to remain accurate throughout the entire session.

Anthropic has not published specific benchmark scores or resolution metrics for Opus 4.7. The improvements are described qualitatively in official documentation. The token context window (the maximum amount of text the model can hold in memory during a session, measured in tokens — roughly 0.75 words each) remains at 1,000,000 tokens, unchanged from Opus 4.6 and Sonnet 4.6. Pricing and access tiers also appear unchanged from the prior model.

Opus 4 Is Gone — And That Is Disruptive for Many Teams

The bigger headline for many teams is not what was added — it is what was taken away. Both Claude Opus 4 and Claude Opus 4.1 have been deprecated (permanently shut down) from Claude and Claude Code as of January 2026. The models are no longer accessible, including through the programmatic interface (the connection point that lets developers integrate Claude into their own applications and automated workflows).

For teams running Opus 4 in production — through automated pipelines, custom integrations, or Claude Code — this is a hard cutoff. Any code that specifically references the old model IDs will now return errors in live systems:

# These model IDs are deprecated and will fail:
# claude-opus-4        (removed January 2026)
# claude-opus-4-1      (removed January 2026)

# Update your code to the current model:
# claude-opus-4-7      (launched April 16, 2026)

Enterprise teams in regulated industries — healthcare, finance, legal — that deliberately stayed on older model versions for compliance stability should audit their integrations now, before unexpected failures surface in live systems. The upgrade path is straightforward, but the audit step cannot wait.

A Model Every 6–8 Weeks: What Anthropic's Cadence Means for Your Stack

Opus 4.7 is the latest step in an aggressive release schedule Anthropic has been running since late 2025. Here is the full model timeline:

• September 2025 — Sonnet 4.5 launched as the fastest model for agent and coding tasks
• October 2025 — Haiku 4.5 matched Sonnet 4 coding and agent performance at lower cost
• November 2025 — Opus 4.5 shipped; Skills marketplace launched with partner-built integrations
• January 2026 — Enterprise plans became HIPAA-ready; Opus 4 and 4.1 deprecated
• February 17, 2026 — Sonnet 4.6 launched with a 1,000,000-token context window (enough to process roughly 750 pages of text in a single session — equivalent to about seven full-length novels)
• March 2, 2026 — Memory feature rolled out to all free users
• April 9, 2026 — Cowork expanded to general availability on macOS and Windows
• April 16, 2026 — Opus 4.7 goes live as the new capability leader

At roughly one major release every 6–8 weeks, Anthropic is operating at a pace closer to software product releases than traditional AI research cycles. The practical consequence for teams building on Claude: treat model version management (tracking which Claude model version your code uses and migrating when older versions retire) as an ongoing engineering task — not a one-time setup decision.

Other Features That Shipped in This Release Wave

Cowork Is Now Open to Enterprise Teams

Claude Cowork — Anthropic's desktop agentic workspace (a desktop app that lets Claude complete multi-step tasks directly on your machine, without uploading your files to a remote server) — reached general availability on April 9, 2026 for both macOS and Windows. The enterprise-focused update added four key capabilities:

• Analytics access — programmatic (code-accessible) visibility into usage data across your organization
• OpenTelemetry monitoring — integration with the open-standard protocol (an industry-wide specification for tracking application performance and system health) used by enterprise IT teams
• Role-based access controls — admins can specify which employees can access which Cowork capabilities
• Usage analytics dashboard — team-wide visibility into AI usage across projects and departments

These additions position Cowork as infrastructure for coordinating AI work across entire organizations — not just a personal productivity add-on. It moves Anthropic into direct competition with Microsoft Copilot for Teams and Google Workspace AI at the enterprise level.

Healthcare Teams Can Now Use Claude Legally for Patient Data

Since January 2026, Anthropic's Enterprise plans are HIPAA-ready — compliant with the Health Insurance Portability and Accountability Act (the US federal law governing how patient data must be stored, processed, and shared). This allows Enterprise customers to process PHI (Protected Health Information — patient records, medical histories, test results, and clinical notes) through Claude without violating federal healthcare privacy regulations.

This compliance gap was a hard blocker for US healthcare organizations wanting to use AI for clinical documentation, diagnosis support, or patient communication. The unlock applies only to Enterprise plan subscribers — not Pro or Max plan users. Interested organizations need a signed Business Associate Agreement with Anthropic before processing patient data.

Who Should Act — and What to Do Right Now

Here is a practical breakdown by user type, ordered by urgency:

• Developers with code referencing Opus 4 or Opus 4.1: Audit immediately. Search your codebase for claude-opus-4 and claude-opus-4-1 and update to claude-opus-4-7. This is the only group facing a live failure risk today.
• Software engineering teams on long Claude Code sessions: Upgrade to Opus 4.7 — it is specifically built for this use case. The vision improvements also benefit teams reviewing UI screenshots, mockups, or architecture diagrams during code reviews.
• Teams currently on Sonnet 4.6: No urgent action needed. Sonnet remains the better cost-per-quality balance for document, email, and general analysis tasks. Switch to Opus 4.7 only when tasks require maximum reasoning depth over many steps.
• US healthcare organizations on Enterprise plans: HIPAA compliance is now live. Confirm your Business Associate Agreement with Anthropic is signed before processing any patient information through Claude.
• Free plan users: Memory is already active on your account as of March 2. Opus 4.7 is now the top available model, but free-tier access is limited — the Opus 4.7 upgrade matters most for Pro ($20/month) and Max ($100/month) subscribers.

You can access Claude Opus 4.7 right now at claude.ai — no installation required for web users. Claude Code and Claude Desktop update automatically. If you use Claude inside automated workflows or pipelines, audit your model version references today before a deprecated ID causes a live failure.

Related Content — Get Started with Claude | AI Automation Guides | More AI News