Claude Code v2.1.73: Model Overrides, Stability Fixes, and Performance Improvements

What's New in Claude Code v2.1.73
Claude Code v2.1.73 introduces practical improvements for developers using AI coding agents, with a focus on stability, configuration flexibility, and cross-platform compatibility.
Key Features and Fixes
- modelOverrides setting: Maps model picker entries to custom provider model IDs, such as Bedrock inference profile ARNs.
- SSL certificate guidance: Added actionable help when OAuth login or connectivity checks fail due to SSL certificate errors, particularly useful for corporate proxies or when using
NODE_EXTRA_CA_CERTS. - Critical stability fixes: Resolved freezes and 100% CPU loops triggered by permission prompts for complex bash commands.
- Deadlock prevention: Fixed a deadlock that could freeze Claude Code when many skill files changed simultaneously, such as during
git pullin repositories with large.claude/skills/directories. - Bash output preservation: Fixed Bash tool output being lost when running multiple Claude Code sessions in the same project directory.
- Model version consistency: Fixed subagents with
model: opus / sonnet / haikubeing silently downgraded to older model versions on Bedrock, Vertex, and Microsoft Foundry. - Process cleanup: Fixed background bash processes spawned by subagents not being cleaned up when the agent exits.
- Command availability: Fixed
/loopnot being available on Bedrock/Vertex/Foundry and when telemetry was disabled. - Session handling: Fixed SessionStart hooks firing twice when resuming a session via
--resumeor--continue. - JSON-output hooks: Fixed injection of no-op system-reminder messages into the model's context on every turn.
- Voice mode stability: Fixed session corruption when a slow connection overlaps a new recording.
- Linux compatibility: Fixed Linux sandbox failing to start with "ripgrep (rg) not found" on native builds, and Linux native modules not loading on Amazon Linux 2 and other glibc 2.26 systems.
- Remote Control fixes: Fixed "media_type: Field required" API error when receiving images via Remote Control.
- Windows compatibility: Fixed
/heapdumpfailing on Windows withEEXISTerror when the Desktop folder already exists. - VSCode proxy support: Fixed HTTP 400 errors for users behind proxies or on Bedrock/Vertex with Claude 4.5 models.
Improvements and Changes
- Up arrow behavior: Improved to restore interrupted prompts and rewind conversations in one step after interrupting Claude.
- IDE detection: Improved startup speed for IDE detection.
- Clipboard performance: Improved image pasting performance on macOS.
- Command behavior: Improved
/effortto work while Claude is responding, matching/modelbehavior. - Voice mode reliability: Improved to automatically retry transient connection failures during rapid push-to-talk re-press.
- Remote Control prompts: Improved spawn mode selection with better context.
- Default model update: Changed default Opus model on Bedrock, Vertex, and Microsoft Foundry to Opus 4.6 (was Opus 4.1).
- Command deprecation: Deprecated
/output-stylecommand — use/configinstead. Output style is now fixed at session start for better prompt caching.
This release addresses multiple edge cases that developers encounter in enterprise environments, particularly around proxy configurations, multi-session workflows, and cross-platform compatibility.
📖 Read the full source: GitHub Claude-Code
👀 See Also

Anthropic Acquires Stainless for $300M+ — Now Owns Dominant MCP Server Generator
Anthropic bought SDK generator Stainless for $300M+. Stainless generates most production MCP servers from OpenAPI specs. The hosted product is winding down; new signups stopped Monday.

Liquid AI releases LFM2.5-350M model for agentic loops
Liquid AI released LFM2.5-350M, a 350M parameter model trained for reliable data extraction and tool use. It's under 500MB when quantized and outperforms larger models like Qwen3.5-0.8B in most benchmarks while being faster and more memory efficient.

Anthropic's Business Strategy: API Revenue Drives Consumer Tier Limitations
Anthropic's consumer subscription tiers operate at a loss, subsidized to build AI mindshare, while their API business generates revenue. The $20 Pro tier is intentionally limited to filter users toward higher-value Max subscriptions.

Medvi's $1.8B AI Company Claims Face Scrutiny Over Legal and Ethical Issues
Gary Marcus critiques the viral story of Medvi, a claimed $1.8B AI company built by one person in two months, highlighting class action lawsuits for spam violations and questions about revenue reporting and compliance.