Claude Platform Updates: New Models, Tools, and Cloud Support
Anthropic has released Claude Fable 5 and Claude Opus 4.8, offering enhanced capabilities like 1M token context windows and improved reasoning. New tool versions for code execution and web search, along with expanded support for Claude Platform on AWS, cater to developers and researchers. The release also includes deprecations of older models like Sonnet 4 and Opus 4, signaling a push towards newer, more powerful versions.
- →Retirement of Claude Sonnet 4 and Opus 4 models
- →Claude Platform on AWS now supports self-hosted sandbox work endpoints
- →Launch of Claude Fable 5 and Claude Mythos 5
- →Code execution tool updated with time limit disclosure
- →Web search and fetch tools gain response inclusion parameter
Deprecations (1) ›
- Retirement of Claude Sonnet 4 and Opus 4 models
Claude Sonnet 4 (claude-sonnet-4-20250514) and Claude Opus 4 (claude-opus-4-20250514) models have been retired. All requests to these models will now result in an error, with users recommended to upgrade to Claude Sonnet 4.6 and Claude Opus 4.8 respectively. Researchers can request continued access via the External Researcher Access Program.
Features (2) ›
- Claude Platform on AWS now supports self-hosted sandbox work endpoints
The GET /v1/environments/{id}/work endpoint, used for listing pending work for self-hosted sandboxes, is now available on Claude Platform on AWS. Authorization for this action is managed through IAM actions for Claude Platform on AWS.
- Launch of Claude Fable 5 and Claude Mythos 5
Anthropic has launched Claude Fable 5 (claude-fable-5), its most capable widely released model, and Claude Mythos 5 (claude-mythos-5) for Project Glasswing participants. Both models feature a default 1M token context window, 128k max output tokens, and adaptive thinking. Introductory materials for capabilities and API changes are available.
Enhancements (5) ›
- Code execution tool updated with time limit disclosure
The code execution tool now supports code_execution_20260521, which includes the 90-second per-cell execution time limit in its description. This allows Claude to better manage long-running cells without requiring a beta header.
- Web search and fetch tools gain response inclusion parameter
The web search and web fetch tools now support web_search_20260318 and web_fetch_20260318, respectively. These updates add a response_inclusion parameter to drop consumed result blocks from API responses for agentic workflows, without needing a beta header.
- New tokenizer for Claude Fable 5 and Mythos 5
Claude Fable 5 and Claude Mythos 5 utilize the tokenizer introduced with Claude Opus 4.7, resulting in approximately 30% more tokens for the same text compared to earlier models. A token counting API with model: "claude-fable-5" is available for measuring prompts under the new tokenizer.
- Claude Fable 5 safety classifiers and refusal handling
Claude Fable 5 incorporates safety classifiers for requests and response generation. Refused requests return stop_reason: "refusal" and are not billed if no output is generated. An opt-in fallback parameter is in beta for re-running refused requests on other models.
- Refusal response category updated for Claude Fable 5
The stop_details.category field on refusal responses for Claude Fable 5 now includes "reasoning_extraction" for requests blocked under Anthropic's Terms of Service restrictions on reverse engineering. Existing categories "cyber" and "bio" remain unchanged.
https://platform.claude.com/docs/en/release-notes/overview