ReleaseBytes
anthropic Claude Platform Release Notes ·

Anthropic Launches Claude Opus 5.5 Model, Introduces New API Capabilities

aipreviewengineer
feature breaking

Anthropic has launched Claude Opus 5.5, an advanced model designed for long-running agentic coding and knowledge work. It features a larger context window and always-on adaptive thinking, alongside new API capabilities like Fast mode and inline tool definitions. This model aims to enhance complex automated workflows and offers developers more flexible ways to manage tools and model behavior. Developers using the Claude API across various cloud platforms will benefit, though some API changes introduce breaking changes for existing integrations, with specific features available in research preview or beta.

  • →Claude Opus 5.5 Model Launch
  • →API Breaking Changes for Thinking and Tool Choice
  • →Computer Use Toolset Update for Claude Opus 5.5
  • →Fast Mode (Research Preview) for Claude Opus 5.5
  • →Inline Tool Definitions in Mid-Conversation System Messages (Beta)
Breaking changes (2) ›
  • API Breaking Changes for Thinking and Tool Choice

    On Claude Opus 5.5, attempts to enable or disable `thinking` via API parameters (`thinking: {"type": "disabled"}` or `thinking: {"type": "enabled", ...}`) will now return a 400 error. Developers should omit the `thinking` field and control thinking depth with the `effort` parameter. Additionally, `tool_choice` types `any` and `tool` now return a 400 error, requiring the use of `auto` with strict tool use.

  • Computer Use Toolset Update for Claude Opus 5.5

    For Claude API and Google Cloud users, computer use on Claude Opus 5.5 now requires the `computer_toolset_20260801` toolset, with the earlier `computer_20251124` tool returning a 400 error. However, on Amazon Bedrock, the `computer_20251124` tool continues to function as before.

Features (3) ›
  • Claude Opus 5.5 Model Launch

    Anthropic has launched Claude Opus 5.5 (`claude-opus-5-5`), a new model for long-running agentic coding and knowledge work. It features a 1M token context window by default, 128k max output tokens, and always-on adaptive thinking, priced at $4/$20 USD per MTok. The model is available on the Claude API, Amazon Bedrock, Claude Platform on AWS, Claude on Google Cloud, and Claude in Microsoft Foundry.

  • Fast Mode (Research Preview) for Claude Opus 5.5

    A new "Fast mode" is now available as a research preview for Claude Opus 5.5 on the Claude API. This mode is designed to offer quicker response times for certain use cases.

  • Inline Tool Definitions in Mid-Conversation System Messages (Beta)

    Tools can now be defined directly within a mid-conversation system message using a `tool_addition` block, available in beta on the Claude API with the `inline-tools-2026-09-15` beta header. This capability allows for dynamic addition, schema changes, or version updates for tools without requiring modifications to the main `tools` array or invalidating the prompt cache.

Read the original announcement →

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

Related releases