What's new in Claude

Session titles now generate in your conversation language, credential caching improved for Bedrock, and fixes for background sessions, Remote Control, and permission handling (including model redirect vulnerability where /fast could switch to blocked models).

Added enforceAvailableModels managed setting to constrain the Default model to the allowlist and prevent settings from widening model restrictions — useful for org-wide model governance.

Added wheelScrollAccelerationEnabled setting to disable mouse-wheel scroll acceleration, fixed /model picker display issues, and corrected Bedrock GovCloud region inference.

Claude Opus 4.8 is now the default model for Max, Team Premium, Enterprise pay-as-you-go, and API accounts — high effort on by default, with /effort xhigh for the hardest tasks.

Dynamic workflows let Claude write and execute a script that orchestrates dozens to hundreds of subagents — replacing one-shot prompts with structured multi-agent plans.

Security-guidance plugin reviews Claude's code changes for vulnerabilities in real time as it works — think of it as a live security co-pilot running alongside every edit.

Fast mode on Opus 4.8 is now available at $10/$50 per MTok input/output — the same speed tier, now on the most capable generally available model.

The GET /v1/environments/{id}/work endpoint for self-hosted sandboxes is now available on Claude Platform on AWS, with the GetEnvironment IAM action authorizing it.

Claude Fable 5 (claude-fable-5) launched as Anthropic's most capable widely released model, with a 1M token context window, 128k max output, always-on adaptive thinking, and safety classifiers that return stop_reason: "refusal" (non-billed) when a request is declined.

Claude Mythos 5 (claude-mythos-5) launches for Project Glasswing participants — same 1M context + 128k output, sitting above Opus in both capability and price.

Claude Managed Agents now supports scheduled deployments (cron-based sessions) and vault environment variable credentials, so you can securely inject secrets into agent sandboxes without a custom scheduler.

Claude Opus 4.1 (claude-opus-4-1-20250805) deprecated — retire date is August 5, 2026. Anthropic recommends migrating to Claude Opus 4.8.

Claude Code weekly limits raised by 50% through July 13 for Pro, Max, Team, and Enterprise users. Also announced: starting June 15, paid-plan users can claim a dedicated monthly credit for programmatic/Agent SDK usage.

Announced Claude Fable 5 and new Managed Agents features including scheduled deployments, multiagent orchestration, and Dreaming — a scheduled process that reviews past agent sessions and curates memory so agents improve between runs.

Boris Cherny marked one year of Claude Code GA: "I haven't written a line of code by hand in eight months. Some mornings I coordinate tens of thousands of agents." Code output at Anthropic is up 8x since the start of 2026.

Thariq Shihipar asked Anthropic teammates how they stay in the loop with Claude — shared a favorite tip from a colleague about tracking agent work across sessions. Quiet since then while team responded to the June 2 service outage.