Claude Code v2.1.126, 새로운 기능 업데이트 소식
Claude Code v2.1.126 has been released with a new Gateway Model Picker and project purge commands, while the simultaneous launch of Claude Opus 4.7 is strengthening the developer platform. Despite some community chatter about usage limits and model quality, developers should update now using `claude update` to test the new `/model` and `claude project purge` features.
Claude Code News Curation — 2026-05-11
🔥 Headline for This Cycle
Claude Code v2.1.126 Released — Gateway Model Picker & Project Purge
Claude Code v2.1.126 was released on May 10, 2026 (local time). The highlight is that the /model picker now dynamically fetches model lists from an Anthropic-compatible gateway's /v1/models endpoint when ANTHROPIC_BASE_URL is configured. This allows enterprise developers using internal proxies or third-party gateways to see available models in real-time without manual setup. Additionally, the claude project purge [path] command was added to clean up project context directly from the CLI.
Claude Opus 4.7 Documentation Update
- Announcement Date: 2026-05-10 (~19 hours ago)
- Details: The Anthropic developer platform documentation has been updated with new features, major changes, and behavioral updates for Claude Opus 4.7. Claude Design (a tool for visual output collaboration) was also launched as a new product from Anthropic Labs alongside Opus 4.7.
- Developer Impact: Teams using the Opus 4.7 API should carefully review behavioral changes and consider adopting Claude Design for workflows involving slides, prototypes, and other visual assets.
Model Overview Documentation Updated
- Announcement Date: 2026-05-10 (~1 day ago)
- Details: The official Anthropic model overview page has been updated, refreshing performance comparisons and recommended use cases for the entire Claude family, including Opus 4.7.
- Developer Impact: A great opportunity to re-evaluate which Claude model to choose for new projects, especially for teams optimizing for cost-to-performance.
Increased Claude Code Usage Limits via SpaceX Partnership
- Announcement Date: 2026-05-09 (~2 days ago)
- Details: Following the announcement of Anthropic’s computing partnership with the SpaceX Colossus One data center, the 5-hour usage limits for Claude Code Pro and Max users have been doubled. Team and Enterprise plans also receive these benefits.
- Developer Impact: This means a direct productivity boost for heavy Claude Code users who previously hit limits while working on large codebases. It applies automatically with no configuration changes required.
🛠 Claude Code Releases & GitHub Trends (2 items)
v2.1.126 — Gateway Integration & Project Purge
- Version/Date: v2.1.126 / 2026-05-10 (~2 days ago)
- Changes:
/modelpicker now dynamically integrates with theANTHROPIC_BASE_URLgateway's/v1/modelsendpoint.- Added
claude project purge [path]command for CLI-based project context cleanup. - Bug fix: Resolved an issue in full-screen mode (for terminals supporting DEC 2026 like iTerm2, Ghostty) where messages were displayed in two locations when scrolling up.
- Upgrade Guide: Run
npm update -g @anthropic-ai/claude-codeorclaude update. Gateway users should check available models via the/modelcommand after setting theANTHROPIC_BASE_URLenvironment variable.
v2.1.92 — Full-screen Scrolling Bug Fix
- Version/Date: v2.1.92 / 2026-05-09 (1 day ago)
- Changes:
- Fixed a rendering bug where scrolling up in full-screen mode caused duplicate message output.
- General stability improvements.
- Upgrade Guide: You can upgrade directly to v2.1.126 from v2.1.92 with a single
claude update.
📰 Tech Media Coverage (2 items)
"Claude Code Usage Limits Doubled — The Meaning of the SpaceX Computing Deal" (Ars Technica)
- Source / Date: Ars Technica, 2026-05-07
- Summary: Ars Technica highlights that the SpaceX partnership is more than just an infrastructure deal, showing Anthropic's rapid supply chain expansion following Microsoft and Amazon. The doubling of the 5-hour usage limit for Claude Code Pro/Max is the key takeaway, signaling a "supply-led growth" strategy to provide more computing power to premium subscribers.
"Claude Code Limits Doubled: SpaceX Expansion Eases Usage Caps" (knightli.com)
- Source / Date: knightli.com, 2026-05-09
- Summary: A deep dive into the May 2026 increases to Claude Code and Claude API limits, including a table breaking down benefits by Pro, Max, Team, and Enterprise plans. It explains the technical background of how the SpaceX computing partnership facilitates these increased limits.
💬 Developer Community Reactions (3 items)
"Update on Claude Code Quality Concerns" (HN)
- Source / Score: Hacker News, hn.item?id=47878905 (2 weeks ago)
- Main Point: Users reported issues with resuming context after sessions were left idle for hours or days. There was significant confusion regarding system prompt changes and default "Thinking" level adjustments. The community appreciated Anthropic's direct response to these quality reports.
"Canceling Claude Subscription: Token Issues, Quality Drops, Poor Support" (HN)
- Source / Score: Hacker News, hn.item?id=47892019 (2 weeks ago)
- Main Point: Some users shared they canceled their subscriptions due to code generation quality issues (missing requirements, redundant code, unnecessary data wrapping). Conversely, many shared workarounds, such as switching to Claude Sonnet, and advised using local models or documentation to save tokens.
"Claude Code Excluded for New $20 Pro Subscribers" (HN)
- Source / Score: Hacker News, hn.item?id=47855832 (2 weeks ago)
- Main Point: Controversy erupted over reports that Claude Code was removed for new $20 Pro plan subscribers. Debate continues over whether this is a move to limit "token-heavy" users or if it ties into the recent usage limit increases.
🔍 Comparison & Analysis
This cycle, Claude Code gained attention from two distinct angles. Positive: The doubled usage limits and v2.1.126’s Gateway Model Picker support clearly target the enterprise market, making Claude Code more flexible than Cursor or GitHub Copilot for those using internal proxies. Negative: Ongoing concerns on Hacker News regarding code quality and subscription policy shifts suggest that in the individual developer market, Claude Code still faces questions regarding price and stability compared to the $20 Cursor Pro. While Claude Code's strength lies in terminal-native agentic execution and long context retention, consistency remains an area for improvement.
🧭 Practical Tips for Developers (3 items)
-
Use the Gateway Model Picker: Set the
ANTHROPIC_BASE_URL=https://your-gateway.example.comenvironment variable and run/modelin a session to automatically fetch available models. Essential for users behind corporate proxies. -
Clean Project Context: As projects grow, accumulated context can impact performance. Periodically run
claude project purge [path]to maintain response speed and accuracy. -
Switch to Sonnet When Hitting Limits: A community-shared workaround—if Claude Code feels restricted or quality dips, switch to Claude Sonnet using the
/modelcommand for better results.
👀 Watch for Next Cycle
- Claude Managed Agents 'Dreaming' Feature: Reports suggest a "dreaming" feature for Claude Managed Agents may soon integrate with Claude Code.
- Pro Plan Policy Clarification: Potential for an official announcement regarding the $20 Pro plan policy and migration guides to the Max plan.
- v2.1.x Minor Release Trend: Expect continued improvements in terminal rendering and multi-file editing as the team maintains a fast-paced minor release cycle.
📌 This Week's Action Items
- Update Now: Run
claude updateornpm update -g @anthropic-ai/claude-code, then checkclaude --versionto ensure you are on v2.1.126 or higher. - Verify Usage Limits: If you are a Pro, Max, or Team user, your 5-hour limit has automatically doubled. Check your plan status via
claude auth status.
This content was collected, curated, and summarized entirely by AI — including how and what to gather. It may contain inaccuracies. Crew does not guarantee the accuracy of any information presented here. Always verify facts on your own before acting on them. Crew assumes no legal liability for any consequences arising from reliance on this content.
Powered by
