What's NewApril 2026· 8 min read
What's New in Claude — April 2026 Update
Claude Opus 4.7 just dropped. Claude Design launched. Amazon committed $25B. Extended thinking matured. A lot happened in 2026 — here's everything worth knowing.
Anthropic has been moving at a breakneck pace in 2026. In just four months we've seen a new flagship model, a visual design product, a cybersecurity-specialized model in private preview, and the largest investment deal in the company's history. This is the full update.
Model Lineup — April 2026
The naming convention has evolved. We're now in the 4.x generation across all tiers.
👑
Opus 4.7
Flagship · Agentic · Vision
LATEST
$5 / $25 per MTok
⚡
Sonnet 4.6
Balanced · Fast · Production
$3 / $15 per MTok
🚀
Haiku 4.5
Fastest · High-volume · Cheap
$1 / $5 per MTok
👑
Claude Opus 4.7 APRIL 16, 2026
The new flagship. Step-change improvement for agentic coding, long-running tasks, and professional document creation. Enhanced vision with 3.75× higher resolution (up to 2,576px). Better instruction following and output verification. Supports 1M token context and up to 128k output tokens. New tokenizer — be aware this affects your token counts if you're migrating.
API: claude-opus-4-7Knowledge cutoff: Jan 2026
⚡
Claude Sonnet 4.6
The workhorse for production applications. 1M token context window now fully available across Max, Team, and Enterprise plans. Adaptive extended thinking built in. Strong across coding, computer use, and long-context reasoning. The right choice for most API workloads.
API: claude-sonnet-4-6Knowledge cutoff: Aug 2025
🚀
Claude Haiku 4.5
Near-frontier intelligence at the lowest cost. Best for high-volume classification, summarization, real-time responses, and anything latency-sensitive. 200k context, 64k output, adaptive thinking available.
API: claude-haiku-4-5-20251001$1/$5 per MTok
🔒
Claude Mythos PRIVATE PREVIEW
Announced April 7, 2026 as Project Glasswing. Specialized for defensive cybersecurity — vulnerability detection, threat analysis, and secure code review. Described as Anthropic's most capable model to date. Currently limited to 11 organizations by invite only. Not available via self-serve API.
Big Features in 2026
💭
Extended Thinking — Now Production-Grade
Thinking has matured significantly. Three key upgrades: Summarized Thinking returns a summary of Claude's reasoning rather than the full chain (privacy + speed win); Adaptive Thinking lets Claude decide when and how much to think based on task complexity — no more guessing the right budget; Interleaved Thinking lets Claude reason between tool calls in agentic workflows, not just before the first response.
GA · No Beta Header Needed
🎨
Claude Design LAUNCHED APR 17
A new product from Anthropic Labs. Powered by Opus 4.7, it generates visual designs, prototypes, slides, and one-pagers from natural language. Available to Pro, Max, Team, and Enterprise subscribers. Currently a research preview — think of it as Claude's answer to the design-from-prompt workflow.
⌨️
Claude Code — Major 2026 Upgrades
Claude Code has evolved into a full development environment. Ultraplan brings cloud-based planning with a web editor for review before remote execution. The new Monitor tool streams background events in real time for live log tailing. Voice input now supports 20 languages including Russian, Polish, Turkish, and Dutch. Write tool is 60% faster. Named sub-agents, team collaboration features, and PID namespace isolation for security.
📐
1M Token Context — Widely Available
As of March 2026, the full 1M token context window (~750,000 words) is available to all Max, Team, and Enterprise users on Sonnet 4.6 and Opus 4.x. That's entire codebases, legal document sets, or research corpora in a single context. Batch API also now supports up to 300k output tokens via the output-300k-2026-03-24 beta header.
🌍
Available Everywhere
All Claude models are now available across Claude API, Amazon Bedrock, Google Vertex AI, and Microsoft Azure AI Foundry. Deploy wherever your infrastructure lives — no more platform lock-in for access to the latest models.
2026 Timeline — Key Dates
April 20, 2026
Amazon commits $25B to Anthropic
$5B immediate injection + $100B in AWS cloud commitments over 10 years. Largest investment in Anthropic's history by a significant margin.
April 17, 2026
Claude Design launches as research preview
Visual design product powered by Opus 4.7 — generate slides, prototypes, and one-pagers from prompts.
April 16, 2026
Claude Opus 4.7 goes GA
New flagship model. Enhanced vision, agentic coding improvements, 128k output, task budgets for full agentic loops.
April 7, 2026
Claude Mythos — Project Glasswing
Cybersecurity-specialized model in private preview. Limited to 11 select organizations.
March 13, 2026
1M context expanded to all users
Full 1M token window available to Max, Team, and Enterprise — no longer gated to select tiers.
February 5, 2026
Claude Opus 4.6 released
Strong jump in coding, planning, and financial analysis. 1M context, extended thinking. Now the second-tier model.
January 2026
Claude Cowork launches
Agentic assistant product with department-specific automation plugins for enterprise workflows.
What to Watch Next
- Claude Mythos public access — the cybersecurity model will likely expand beyond private preview in the coming months
- Claude Design maturation — research preview → full product; could compete directly with design-from-prompt tools
- TPU compute expansion — Google/Broadcom partnership for next-gen TPUs, deployment starting 2027, will unlock further model scale
- Task budgets in Claude Code — Opus 4.7's ability to budget full agentic loops (thinking + tool calls + output) will reshape how complex automations are built
How to Stay Current
🚀 The pace is still accelerating. Two major model releases, a new product category, and a $25B investment in four months. If you're building with Claude, keep one eye on the models overview doc — what's cutting-edge today may be the baseline by summer.
ClaudeAIOpus 4.7Anthropic2026Extended Thinking
👨💻
Mayur Rele
Senior Director, IT & Information Security · Parachute Health
15+ years in DevOps, cloud, and cybersecurity. 700+ research citations. Scientist of the Year 2024.