ReleaseBytes
anthropic Claude Platform Release Notes ·

Claude Platform Updates: New Models, Hooks, and SDK Enhancements

aipreviewengineer
feature patch deprecation

Anthropic has released Claude Opus 5 with a 1M token context window and updated pricing. Key features include new inference hooks for enterprise security, expanded support for Dreams research preview, and mid-conversation tool changes now in beta. These updates affect developers using the Claude API, SDKs, and Console, with some changes requiring migration for continued access to specific features.

  • →Claude Opus 5 and 1M Token Context Window Now Available
  • →Inference Hooks Beta for Enterprise Security
  • →Opus 5 Disabling 'Thinking' Behavior Change
  • →Dreams Research Preview Supports New Models
  • →Mid-conversation Tool Changes Beta Expanded
Breaking changes (1) ›
  • Opus 5 Disabling 'Thinking' Behavior Change

    Disabling 'thinking' in Claude Opus 5 is now restricted to effort levels 'high' or below, with attempts at 'xhigh' or 'max' returning a 400 error. This is a breaking change from Claude Opus 4.8.

Features (2) ›
  • Claude Opus 5 and 1M Token Context Window Now Available

    Claude Opus 5 (claude-opus-5) is launched with a 1 million token context window and 128k max output tokens. It offers 'thinking on' by default and retains the pricing of Claude Opus 4.8, available via API, AWS Bedrock, Google Cloud, and Microsoft Azure.

  • Inference Hooks Beta for Enterprise Security

    Inference hooks are now in beta for Claude Enterprise organizations, allowing AI security servers to vet each prompt before inference. This feature supports signed requests, configurable failure handling, and logs denials in the compliance Activity Feed.

Enhancements (3) ›
  • Dreams Research Preview Supports New Models

    The Dreams research preview now supports Claude Opus 5, Claude Fable 5, and Claude Sonnet 5, expanding the models available for experimental use.

  • Mid-conversation Tool Changes Beta Expanded

    Mid-conversation tool changes are now in beta for Claude Fable 5, Claude Mythos 5, Claude Opus 4.8, and Claude Opus 5, allowing tools to be added or removed during a conversation while preserving the prompt cache.

  • New Fallbacks Parameter Mode and Admin API for Enterprise

    The `fallbacks` parameter now supports a 'default' mode for Anthropic's recommended fallback models. Additionally, the Admin API for Claude Enterprise organizations is in beta for managing users, roles, and groups.

Read the original announcement →

https://platform.claude.com/docs/en/release-notes/overview

Related releases