Claude Platform Updates: New Models, Rate Limits, and Tooling
Anthropic has released Claude Fable 5 and Claude Mythos 5 with large context windows and adaptive thinking, alongside Claude Opus 4.8 as its most capable GA model. Rate limits have been harmonized across API tiers, and support for new code execution and web search tool versions has been added. These updates affect users of the Claude API, SDKs, and Console, particularly those leveraging advanced AI capabilities and seeking improved performance.
- →Launch of Claude Fable 5 and Claude Mythos 5 models
- →Claude Opus 4.8 released as most capable GA model
- →Fast mode deprecated for Claude Opus 4.7, removal scheduled
- →Claude Sonnet 4 and Claude Opus 4 models retired
- →Claude Opus 4.1 model deprecation announced
Deprecations (3) ›
- Fast mode deprecated for Claude Opus 4.7, removal scheduled
Fast mode for Claude Opus 4.7 is deprecated and will be removed on July 24, 2026. Requests to this mode will return an error after removal; migration to fast mode for Claude Opus 4.8 is recommended.
- Claude Sonnet 4 and Claude Opus 4 models retired
The Claude Sonnet 4 (claude-sonnet-4-20250514) and Claude Opus 4 (claude-opus-4-20250514) models have been retired, and all requests to them will now return an error. Users are advised to upgrade to Claude Sonnet 4.6 and Claude Opus 4.8 respectively.
- Claude Opus 4.1 model deprecation announced
The Claude Opus 4.1 model (claude-opus-4-1-20250805) is deprecated, with retirement on the Claude API scheduled for August 5, 2026. Migration to Claude Opus 4.8 is recommended.
Features (2) ›
- Launch of Claude Fable 5 and Claude Mythos 5 models
New models, Claude Fable 5 and Claude Mythos 5, are now available with a 1M token context window, 128k max output tokens, and always-on adaptive thinking. Claude Fable 5 also features safety classifiers with a new 'reasoning_extraction' refusal category and requires 30-day data retention. These models are compatible with the tokenizer introduced with Claude Opus 4.7.
- Claude Opus 4.8 released as most capable GA model
Claude Opus 4.8 is now generally available, supporting a 1M token context window on various platforms, 128k max output tokens, and the same tools and features as Claude Opus 4.7.
Enhancements (3) ›
- Rate limits harmonized and usage tiers consolidated for Claude API
Claude Sonnet and Haiku rate limits now match Claude Opus across all tiers, which have been consolidated into 'Start', 'Build', and 'Scale'. Most organizations will move to a higher tier, with no organization receiving lower limits.
- SDKs updated for new code execution tool version with state persistence
Python, TypeScript, Go, Java, Ruby, PHP, and C# SDKs now support code_execution_20260120, which includes REPL state persistence and is the minimum version for programmatic tool calling. This version is available on several Claude models, including Fable 5 and Opus 4.5 and newer.
- Code execution tool updated for execution time limits and web tools for response filtering
The code execution tool now supports version code_execution_20260521, disclosing a 90-second per-cell execution time limit. Web search and web fetch tools also support versions that add a response_inclusion parameter to filter consumed result blocks.
https://platform.claude.com/docs/en/release-notes/overview