Claude AI Updates: Everything Anthropic Shipped in March-April 2026

Anthropic shipped more in March and April 2026 than most AI companies ship in a year. A new flagship model. A visual design product. A cybersecurity-focused frontier model so capable it triggered a gated research release. Over 30 Claude Code updates in five weeks. And a user backlash over perceived quality drops that made Fortune headlines.

If you build content for AI search engines, optimize for AI visibility, or simply use Claude daily, every one of these updates changes something about how you work. This roundup covers all of them, what shipped, when, and why it matters.

Opus 4.7: the new flagship (April 16)

Claude Opus 4.7 replaced Opus 4.6 as the most capable Claude model on April 16, 2026. The improvements focus on three areas: agentic coding, vision, and task execution control.

Software engineering gains. On Anthropic’s internal 93-task coding benchmark, Opus 4.7 lifted resolution by 13 percent over Opus 4.6, including four tasks that neither Opus 4.6 nor Sonnet 4.6 could solve. Users report being able to hand off their hardest coding work, multi-file refactors, complex debugging, architecture decisions, to Opus 4.7 with confidence.

Vision upgrade. Opus 4.7 is the first Claude model with high-resolution image support. Maximum image resolution increased from 1,568 pixels (1.15 megapixels) to 2,576 pixels on the long edge (3.75 megapixels). This means Claude can now accurately read complex user interfaces, detailed design files, existing websites, and intricate diagrams, a meaningful jump for anyone using Claude for UI review, document analysis, or visual QA.

New control features. Two new mechanisms give users finer control over how Opus 4.7 allocates compute:

xhigh effort level. A new tier between high and max that offers deeper reasoning without the full latency cost of max. Available via /effort in Claude Code and the API.
Task budgets. A new feature that gives Claude a rough estimate of how many tokens to target for a full agentic loop. The model sees a running countdown and uses it to prioritize work and finish gracefully, preventing the common problem of agentic runs that spiral or stall.

Pricing and availability. Same as Opus 4.6: $5 per million input tokens and $25 per million output tokens. Available across Claude.ai, the API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.

For AI SEO practitioners, the vision upgrade matters most. Claude can now parse full web page screenshots with enough resolution to evaluate layout, typography, and content hierarchy, useful for competitive audits and AI visibility assessments.

Claude Design: visual creation without a design tool (April 17)

One day after Opus 4.7, Anthropic launched Claude Design, a new Anthropic Labs product that lets users create visual assets through conversation.

Instead of opening Figma or Canva, users describe what they want in plain language. Claude generates a working visual, prototype, wireframe, slide deck, social asset, one-pager, that can be refined through follow-up conversation, inline comments, direct edits, or custom sliders that Claude builds on the fly.

Key capabilities:

Design system awareness. Claude Design can read a team’s codebase and existing design files to apply consistent brand styling across every project it creates.
Export flexibility. Output exports to PDF, URL, PPTX, or directly to Canva.
High-fidelity prototypes. Powered by Opus 4.7’s upgraded vision, Claude Design can process existing UI screenshots and design files at triple the previous resolution, then generate variations or improvements.

Availability. Research preview for Claude Pro, Max, Team, and Enterprise subscribers.

For content teams, Claude Design is significant because it collapses the gap between ideation and visual output. Blog headers, presentation decks, social media assets, and wireframes that previously required a designer or a separate tool can now be drafted inside the same Claude conversation where the content strategy lives.

Timeline of Anthropic Claude releases — Photo by AISEOShift on AISEOShift

Claude Mythos Preview: the model that leaked (March 26 - April 8)

Claude Mythos is the most unusual Anthropic release in the company’s history. It was not announced, it was discovered.

The leak. On March 26, Fortune reported that a CMS misconfiguration on Anthropic’s website briefly exposed documentation for a model codenamed Mythos. Anthropic confirmed the leak and described Mythos as a “step change in capabilities” compared to existing models.

The benchmarks. When Anthropic officially released Mythos Preview on April 8, the numbers explained the secrecy: SWE-bench 93.9 percent, USAMO 97.6 percent, with dramatically higher scores in coding, academic reasoning, and cybersecurity than any Claude model before it.

The cybersecurity factor. Mythos Preview is capable of identifying and exploiting zero-day vulnerabilities in every major operating system and every major web browser when directed by a user. Anthropic used Mythos to identify thousands of previously unknown vulnerabilities across critical software.

Project Glasswing. Rather than releasing Mythos broadly, Anthropic launched Project Glasswing, an initiative to use Mythos Preview to help secure the world’s most critical software. Anthropic and AWS are prioritizing access for internet-critical companies and open-source maintainers whose software impacts hundreds of millions of users. CrowdStrike became one of the first founding members.

Gated access. Mythos Preview is available on red.anthropic.com for vetted security researchers, on Amazon Bedrock (gated research preview), and on Google Cloud Vertex AI. It is not available to general consumers.

For the AI visibility ecosystem, Mythos matters because it signals a future where frontier models specialize. A model that finds zero-days faster than human security teams will reshape how we think about website security, content integrity, and the trustworthiness signals that AI answer engines evaluate.

Claude Code: 30+ releases in five weeks

Claude Code entered its most intensive iteration cycle between March and April 2026, pushing from v2.1.69 to v2.1.101, over 30 releases in five weeks. The major changes cluster around performance, developer experience, and security.

Performance upgrades

1M context window GA. Opus 4.6 with a 1 million token context window became generally available, with output token limits increased to 64K-128K.
NO_FLICKER rendering engine. A new rendering system that eliminates screen flicker during code generation, making long sessions more comfortable.
Focus View. A new UI mode that highlights the active file and hides noise during complex edits.
60 percent Write tool speedup. File writes became dramatically faster.

Developer experience

/powerup interactive tutorials. New guided tutorials that teach Claude Code features through hands-on exercises rather than documentation.
Auto mode for Max subscribers. When using Opus 4.7, Claude Code can now run autonomously with minimal user intervention, reading files, making changes, running tests, and iterating without pausing for approval at each step.
Session recap. A new feature that summarizes what happened during a session, useful for handoffs and reviewing work done by autonomous agents.
1-hour and forced 5-minute prompt caching. New caching controls that reduce latency and cost on repeated operations.
Computer use in CLI. Research preview of Claude controlling native applications, opening apps, clicking through UI, and verifying changes visually.

Security hardening

PID namespace isolation for sandboxed execution
Credential scrubbing to prevent accidental exposure of secrets
Hardened PowerShell permissions on Windows
Fixes for command injection vulnerabilities

Collaboration features

/team-onboarding guides for setting up new team members
Named sub-agents for organizing complex multi-step tasks
Mobile push notifications for long-running background tasks

Platform and business updates

Beyond models and tools, Anthropic made several structural moves:

Anthropic Institute. Co-founder Jack Clark launched the Anthropic Institute, focused on AI policy research and safety standards.

Claude Partner Network. Anthropic committed $100 million to partner integrations, accelerating Claude’s presence in third-party platforms and enterprise workflows.

Model deprecations. Sonnet 4.0 and Opus 4.0 (the original Claude 4 family) are scheduled for retirement on June 15, 2026. Teams still using these models should migrate to Sonnet 4.6 or Opus 4.7.

Subscription changes. The Max plan launched at $100 per month (5x Pro usage) or $200 per month (20x), targeting power users who need heavy daily throughput. Team plans now start at $25 per user per month with a $150 premium seat option that includes Claude Code access.

Usage growth. Claude Code usage grew 300 percent during Q1 2026, and Anthropic’s revenue increased 5.5x year over year.

The performance controversy

Not everything went smoothly. In mid-April, multiple user communities reported that Claude responses felt less thorough, less creative, and more likely to produce surface-level output, even on paid tiers.

Fortune covered the backlash on April 14, reporting widespread user complaints about “perceived quality degradation” and “lack of transparency” from Anthropic. The controversy connected to changes in default effort levels and compute allocation that Anthropic had not publicly communicated.

Anthropic responded by adjusting effort defaults and improving transparency around how effort levels affect output quality. The episode highlighted a tension inherent in scaling AI services: balancing compute costs against user expectations as subscriber counts grow.

For content professionals who rely on Claude for AI-assisted writing and SEO workflows, the lesson is straightforward: always check your effort level settings and test output quality regularly. The difference between default and high effort can be the difference between thin output and genuinely useful analysis.

What is coming next

Sonnet 4.8. Expected in May 2026, Sonnet 4.8 should bring Opus-level improvements to the faster, cheaper model tier. No official timeline from Anthropic.

Claude Mythos general release. Timeline unclear. Anthropic is still evaluating safety implications before broader access. The gated research preview on Amazon Bedrock and Vertex AI suggests enterprise access will come before consumer access.

Claude Design GA. Currently in research preview. Anthropic has not announced a general availability date, but the rapid iteration on Opus 4.7’s vision capabilities suggests the visual product line is a priority.

What this means for AI SEO

Three takeaways for teams working on AI visibility and answer engine optimization:

Claude is getting better at reading your site. Opus 4.7’s vision upgrade means Claude can now parse full-page screenshots at 3.75 megapixels. If your content looks good visually and is structured for AI extraction, Claude can evaluate both the text and the layout quality.
Agentic workflows will reshape competitive research. Claude Code’s auto mode and computer use preview mean that automated competitive audits, where Claude visits competitor sites, evaluates their content, and reports findings, are now practical at scale.
Quality still depends on effort settings. The performance controversy proved that default AI output quality fluctuates. Teams building content workflows around Claude should pin effort levels and test output quality regularly rather than assuming consistent performance.

Frequently asked questions

What is the latest Claude model in April 2026?

Claude Opus 4.7 is the latest generally available flagship model, released April 16, 2026. Claude Mythos Preview is a more capable model available only through gated research access for vetted security researchers and select enterprise partners.

How much does Claude cost in 2026?

Claude offers a free tier, Pro at $20 per month, Max at $100 to $200 per month, Team at $25 per user per month, and Enterprise at custom pricing. API pricing for Opus 4.7 is $5 per million input tokens and $25 per million output tokens.

What is Claude Mythos?

Claude Mythos is Anthropic’s most capable model, first revealed through an accidental data leak in March 2026 and officially released as a gated research preview in April. It achieves 93.9 percent on SWE-bench and has advanced cybersecurity capabilities. It is not available to general consumers.

What is Claude Design?

Claude Design is an Anthropic Labs product launched April 17, 2026 that lets users create visual assets, prototypes, slide decks, social graphics, wireframes, through natural language conversation. It is powered by Opus 4.7 and available in research preview for Pro, Max, Team, and Enterprise subscribers.

Is Claude better than ChatGPT in 2026?

It depends on the task. Claude Opus 4.7 leads on complex software engineering and long-form reasoning tasks. GPT-5 leads on multimodal versatility and ecosystem breadth. Claude Mythos benchmarks exceed both, but it is not publicly available. For a detailed comparison, see our Claude vs ChatGPT breakdown.