New modelOverrides setting (2.1.73) lets enterprise admins map individual model versions to provider-specific IDs such as Bedrock inference profile ARNs, enabling per-version routing and governance
New autoMemoryDirectory setting (2.1.74) allows configuring a custom directory for auto-memory storage; /context command gains actionable optimization tips
API Tier 4 monthly spend threshold raised dramatically from $5,000 to $200,000
/output-style command deprecated in favor of /config; output style is now locked at session start to enable prompt caching
1-hour prompt caching now supported on Google Cloud's Vertex AI (previously only Claude API and Microsoft Foundry)
Added modelOverrides setting to map model picker entries to custom provider model IDs (e.g., Bedrock inference profile ARNs), enabling per-version routing for enterprise environments
Added actionable guidance when OAuth login or connectivity checks fail due to SSL certificate errors (corporate proxies, NODE_EXTRA_CA_CERTS)
Existing feature improvements
Changed default Opus model on Bedrock, Vertex, and Microsoft Foundry to Opus 4.6 (was Opus 4.1)
Deprecated /output-style command — use /config instead; output style is now fixed at session start for better prompt caching
Improved Up arrow after interrupting Claude — now restores the interrupted prompt and rewinds the conversation in one step
Improved /effort to work while Claude is responding, matching /model behavior
Improved IDE detection speed at startup and clipboard image pasting performance on macOS
Improved voice mode to automatically retry transient connection failures during rapid push-to-talk re-press
Major bug fixes
Fixed freezes and 100% CPU loops triggered by permission prompts for complex bash commands
Fixed a deadlock that could freeze Claude Code when many skill files changed at once (e.g., during git pull in a repo with a large .claude/skills/ directory)
Fixed Bash tool output being lost when running multiple Claude Code sessions in the same project directory
Fixed subagents with model: opus/sonnet/haiku being silently downgraded to older model versions on Bedrock, Vertex, and Microsoft Foundry
Fixed SessionStart hooks firing twice when resuming a session via --resume or --continue
Fixed JSON-output hooks injecting no-op system-reminder messages into the model's context on every turn
Fixed Linux sandbox failing to start with "ripgrep (rg) not found" on native builds
Fixed Linux native modules not loading on Amazon Linux 2 and other glibc 2.26 systems
VSCode: Fixed HTTP 400 errors for users behind proxies or on Bedrock/Vertex with Claude 4.5 models
Added actionable suggestions to /context command — identifies context-heavy tools, memory bloat, and capacity warnings with specific optimization tips
Added autoMemoryDirectory setting to configure a custom directory for auto-memory storage
Existing feature improvements
Changed --plugin-dir so local dev copies now override installed marketplace plugins with the same name (unless force-enabled by managed settings)
Major bug fixes
Fixed memory leak where streaming API response buffers were not released when the generator was terminated early, causing unbounded RSS growth on the Node.js/npm code path
Fixed managed policy ask rules being bypassed by user allow rules or skill allowed-tools
Fixed full model IDs (e.g., claude-opus-4-5) being silently ignored in agent frontmatter model: field and --agents JSON config
Fixed MCP OAuth authentication hanging when the callback port is already in use
Fixed MCP OAuth refresh never prompting for re-auth after the refresh token expires for OAuth servers that return errors with HTTP 200 (e.g., Slack)
Fixed SessionEnd hooks being killed after 1.5s on exit regardless of hook.timeout — now configurable via CLAUDE_CODE_SESSIONEND_HOOKS_TIMEOUT_MS
Fixed Hebrew, Arabic, and other RTL text not rendering correctly in Windows Terminal, conhost, and VS Code integrated terminal
Fixed LSP servers not working on Windows due to malformed file URIs
VSCode: Fixed delete button not working for Untitled sessions
New section added explaining how to use modelOverrides to map multiple model versions within the same family to distinct Bedrock inference profile ARNs. [lines 209-230] [Source]
New section "Override model IDs per version" documenting the modelOverrides setting, which maps individual Anthropic model IDs to provider-specific strings for per-version routing on Bedrock, Vertex, and Foundry. [lines 227-252] [Source]
Instructions for changing output style updated: now uses /config menu instead of the deprecated /output-style command. [lines 35-58] [Source]
Added note that output style is fixed at session start so the system prompt stays stable, enabling prompt caching to reduce latency and cost. [lines 55-58] [Source]
Added modelOverrides to the settings reference table, describing how to map Anthropic model IDs to provider-specific model IDs such as Bedrock inference profile ARNs. [line 158] [Source]
Google Cloud's Vertex AI added to the list of platforms supporting 1-hour prompt caching (previously only Claude API and Microsoft Foundry). [line 75] [Source]