CLAUDE LABJP
FORK — Claude Code 2.1.212 changes what /fork does: it copies your conversation into a new background session with its own row in claude agents, so you can keep working. The old in-session subagent is now /subtaskLIMITS — WebSearch calls are now capped at 200 per session by default, and subagent spawns get the same 200 ceiling, so a runaway search or delegation loop stops on its ownMCPBG — MCP tool calls running past two minutes now move to the background automatically, keeping the session usable. Tune the threshold with CLAUDE_CODE_MCP_AUTO_BACKGROUND_MSPLANFIX — Fixed plan mode auto-running file-modifying Bash commands such as touch and rm without a permission prompt or an SDK canUseTool callbackSONNET5 — Claude Sonnet 5 is running on introductory pricing of $2 per million input tokens and $10 per million output. After August 31 it moves to $3 and $15IPO — Bankers are reportedly lining up investor meetings for Anthropic ahead of a possible public listing as soon as OctoberFORK — Claude Code 2.1.212 changes what /fork does: it copies your conversation into a new background session with its own row in claude agents, so you can keep working. The old in-session subagent is now /subtaskLIMITS — WebSearch calls are now capped at 200 per session by default, and subagent spawns get the same 200 ceiling, so a runaway search or delegation loop stops on its ownMCPBG — MCP tool calls running past two minutes now move to the background automatically, keeping the session usable. Tune the threshold with CLAUDE_CODE_MCP_AUTO_BACKGROUND_MSPLANFIX — Fixed plan mode auto-running file-modifying Bash commands such as touch and rm without a permission prompt or an SDK canUseTool callbackSONNET5 — Claude Sonnet 5 is running on introductory pricing of $2 per million input tokens and $10 per million output. After August 31 it moves to $3 and $15IPO — Bankers are reportedly lining up investor meetings for Anthropic ahead of a possible public listing as soon as October
Articles/Claude.ai
Claude.ai/2026-04-04Beginner

Claude April 2026 Updates: Complete Roundup — /powerup Lessons, API Expansions & Deprecation Deadlines

Everything you need to know about Claude's April 2026 updates: Claude Code /powerup interactive lessons, Batches API 300k output tokens, Claude Haiku 3 retirement (April 19), 1M context beta sunset (April 30), and more.

claude-ai15update6claude-code129api38202616

April 2026 brings a focused wave of Claude updates — from a brand-new interactive learning experience in Claude Code to important deprecation deadlines that could affect your production applications. Here's everything you need to know.

Claude Code Updates

/powerup — Interactive Lessons Are Here

The April 1 release introduced /powerup, an in-session command that walks you through Claude Code features using animated, step-by-step demos. Rather than reading documentation, you learn by doing — right inside your terminal.

# Run inside a Claude Code session
/powerup

After running the command, you'll see a menu of available lessons covering topics like hooks configuration, MCP server setup, and multi-agent workflows. Each lesson is self-paced and interactive, making it ideal for users who are newer to Claude Code or who want to explore features they haven't tried yet. For a deep dive into everything /powerup offers, check out the Claude Code /powerup Command Guide.

Critical Bug Fixes

Several high-impact bugs were resolved in this release cycle:

Rate limit dialog infinite loop (Critical) After hitting a usage limit, the rate-limit options dialog would repeatedly auto-open, eventually crashing the entire session. This is now fixed — a welcome relief for teams running intensive workloads.

--resume causing full prompt-cache miss Users with deferred tools, MCP servers, or custom agents configured were experiencing a complete prompt-cache miss on the very first request after using --resume. This caused slower cold-start performance and higher token costs. The fix makes resume work as expected. See the Claude Code April 2026 Update — Deferred Permission & MCP article for related context.

PostToolUse hook triggering "File content has changed" errors When a PostToolUse hook (e.g., a format-on-save script) rewrote a file between consecutive edits, the next Edit or Write operation would fail with a "File content has changed" error. This is now resolved, which particularly benefits teams using Prettier, ESLint, or similar formatters as hooks.

PreToolUse hook JSON stdout + exit code 2 not blocking tool calls Hooks that wrote JSON to stdout and exited with code 2 were not correctly blocking the tool call they were intended to block. This behavior is now fixed.

Other Additions

  • New CLAUDE_CODE_PLUGIN_KEEP_MARKETPLACE_ON_FAILURE environment variable keeps the existing marketplace cache intact when a git pull fails — useful in offline or constrained network environments.
  • .husky added to the protected directories list in acceptEdits mode.
  • X-Claude-Code-Session-Id header is now included in all API requests, making it easier for proxies to aggregate requests by session without body inspection.

Anthropic API Updates

Batches API Now Supports 300k Output Tokens

The Message Batches API now allows up to 300,000 output tokens per request for Claude Opus 4.6 and Claude Sonnet 4.6. To enable this, include the output-300k-2026-03-24 beta header in your Batches API call.

import anthropic
 
client = anthropic.Anthropic()
 
# Enable 300k output tokens via the Batches API
response = client.beta.messages.batches.create(
    requests=[
        {
            "custom_id": "long-form-001",
            "params": {
                "model": "claude-sonnet-4-6",
                "max_tokens": 300000,
                "messages": [
                    {
                        "role": "user",
                        "content": "Generate a comprehensive technical specification for..."
                    }
                ]
            }
        }
    ],
    betas=["output-300k-2026-03-24"]
)
 
print(response.id)
# Output example: msgbatch_01XFDUDYJgAACcwz5riDLjBX

This is particularly powerful for use cases involving long-form document generation, large-scale code output, and bulk structured data extraction. Combined with the async nature of the Batches API, you can queue up many high-volume jobs cost-effectively.

Models API Now Returns Capability Fields

The GET /v1/models endpoint now returns max_input_tokens, max_tokens, and a capabilities object for each model. This enables applications to dynamically discover model limits at runtime rather than hardcoding them.

import anthropic
 
client = anthropic.Anthropic()
models = client.models.list()
 
for model in models.data:
    print(f"{model.id}: max_input={model.max_input_tokens}, max_output={model.max_tokens}")
    # Example output:
    # claude-sonnet-4-6-20261018: max_input=1000000, max_output=8192
    # claude-opus-4-6-20261018: max_input=1000000, max_output=8192

This is a small but meaningful quality-of-life improvement for developers building multi-model applications.

Web Search & Programmatic Tool Calling Are Now Generally Available

Both the web search tool and programmatic tool calling have exited beta and are now generally available — no beta headers required. If you've been including betas=["tools-2024-05-16"] or similar headers in your requests, you can remove them.

Additionally, web search and web fetch now support dynamic filtering: results are filtered using sandboxed code execution before reaching the context window. This means your model receives only the most relevant content, reducing token usage and improving response quality.

# Before GA (beta header required)
response = client.beta.messages.create(
    model="claude-sonnet-4-6",
    betas=["web-search-2025-03-05"],  # No longer needed
    ...
)
 
# After GA (no beta header)
response = client.messages.create(
    model="claude-sonnet-4-6",
    # No beta header required
    tools=[{"type": "web_search_20250305", "name": "web_search"}],
    messages=[{"role": "user", "content": "What happened in AI today?"}]
)

Deprecation & Migration Deadlines

Claude Haiku 3 Retires April 19, 2026

Claude Haiku 3 will be fully retired on April 19, 2026. If any of your applications currently call claude-haiku-3 (or claude-haiku-3-20240307), they will stop working after that date unless migrated.

Claude Haiku 4.5 is the recommended replacement — it matches or exceeds Haiku 3's speed while delivering meaningfully better output quality at comparable pricing. For step-by-step migration instructions, see the Claude Haiku 3 Deprecation & Migration Guide.

Quick migration:

# Before
model = "claude-haiku-3-20240307"  # Retiring April 19
 
# After
model = "claude-haiku-4-5-20251001"  # Recommended replacement

1M Token Context Beta Ends April 30, 2026

The 1M token context window beta for Claude Sonnet 4.5 and Claude Sonnet 4 ends on April 30, 2026. After that date, the context-1m-2025-08-07 beta header will have no effect on these models — they will revert to their standard context limits.

To continue using 1M token context windows, migrate to Claude Sonnet 4.6 or Claude Opus 4.6, both of which support 1M context at standard pricing with no beta header required.

# Before: Sonnet 4.5 with beta header (stops working April 30)
response = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=4096,
    betas=["context-1m-2025-08-07"],  # Ineffective after April 30
    messages=[...]
)
 
# After: Sonnet 4.6 — no beta header needed
response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=4096,
    # No betas parameter needed
    messages=[...]
)

If you're building applications that rely on long context windows, now is the time to test and validate the migration.

Agent Skills on claude.ai

Claude.ai now includes Agent Skills — pre-built, task-specific skills available in conversations with Extended Thinking enabled. The initial set focuses on document creation and editing workflows, allowing Claude to take multi-step actions within a single conversation without manual back-and-forth.

This marks a step toward a more capable, agentic experience directly in the web interface, complementing the automation already possible in Claude Code and Cowork environments.

Data Residency Controls

The API now supports an inference_geo parameter, giving you control over which geographic region model inference runs in. US-only inference is available at 1.1x pricing for models released after February 1, 2026.

response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    inference_geo="us",  # Inference runs only in US regions
    messages=[{"role": "user", "content": "Process this compliance-sensitive document..."}]
)

This feature is especially relevant for healthcare, financial services, and government applications that require data to remain within specific geographic boundaries.

Looking back

April 2026's Claude updates cover a lot of ground — developer experience, API capabilities, and critical migration deadlines all in one wave:

  • /powerup makes learning Claude Code more accessible with in-session animated lessons
  • Batches API can now handle 300k output tokens per request for large-scale generation tasks
  • Web search and programmatic tool calling are fully GA — remove those beta headers
  • Haiku 3 retires April 19 and 1M context beta ends April 30 — don't miss these dates
  • Agent Skills and data residency controls open new doors for enterprise and compliance use cases

If you're looking to get the most out of the models powering all of this, the Claude Sonnet 4.6 Complete Mastery Guide is a great next step.

Share

Thank You for Reading

Claude Lab is ad-free, supported entirely by members like you. We publish practical guides daily with implementation code, benchmarks, and production-ready patterns. If you've found it useful, we'd love to have you on board.

  • Copy-paste ready implementation code
  • New advanced guides published daily
  • $5/mo or $10 for lifetime access
View Membership →

If you found this article helpful, a small tip ($1.50) would mean a lot to us. Your support helps keep this site ad-free and covers server and hosting costs.

Related Articles

Claude.ai2026-05-04
Anthropic IPO 2026: A Playbook for Developers and Investors Reading the Same News Differently
Anthropic IPO coverage in 2026 is everywhere, but almost all of it is investor-facing. This playbook integrates the investor lens with the developer lens — what changes for API pricing, roadmap cadence, competitive dynamics, and how to prepare your own project.
Claude.ai2026-03-26
How to Summarize Text Effectively with Claude AI — A from Prompt Design to API Integration
Master text summarization with Claude AI. Learn prompt design patterns for executive summaries, meeting notes, research papers, and PDFs, plus Python and TypeScript API implementations for automated workflows.
Claude.ai2026-06-12
Three Days of Running Claude Fable 5 Side by Side with Opus 4.8 — Settling My Model Split Before June 15
I ran Claude Fable 5 alongside Opus 4.8 on the same iOS maintenance tasks during the free introduction window. Where the new model clearly pulls ahead, where it does not, and how I split the two before the June 15 billing change.
📚RECOMMENDED BOOKS
Build a Large Language Model (From Scratch)
Sebastian Raschka
LLM Dev
Prompt Engineering for LLMs
Berryman & Ziegler
Prompting
AI Engineering
Chip Huyen
AI Eng
* Contains affiliate links
See all →