Anthropic SDK v0.108.0 Adds Claude Mythos-5 and Fable-5 with Fallback Support
Anthropic released Python SDK v0.108.0 on June 9, 2026, adding support for two new Claude models: Claude Mythos-5 and Claude Fable-5. The update includes both server-side and client-side fallback middleware for handling model refusals.
Anthropic SDK v0.108.0 Adds Claude Mythos-5 and Fable-5 with Fallback Support
Anthropic released version 0.108.0 of its Python SDK on June 9, 2026, introducing support for two previously unannounced Claude models: Claude Mythos-5 and Claude Fable-5.
New Models
The SDK update references two new model identifiers:
- Claude Mythos-5
- Claude Fable-5
No technical specifications, pricing, or capability details have been disclosed for either model. The naming convention breaks from Anthropic's established pattern of Opus, Sonnet, and Haiku tiers.
Fallback Mechanisms
The primary feature of this release is fallback support for handling model refusals:
Server-side fallbacks: When Claude Mythos-5 or Fable-5 refuse a request, the system can automatically route to an alternative model on Anthropic's servers.
Client-side fallbacks: New middleware enables developers to implement fallback logic in their applications for API providers that don't support server-side fallbacks.
The fallback functionality specifically targets refusal scenarios—instances where models decline to respond to certain prompts due to safety filters or content policies.
Implementation Details
The update (commit 6b76649) implements both fallback types as part of the SDK's core functionality. Developers using the Python SDK can now configure fallback chains to maintain service continuity when models refuse requests.
Version 0.108.0 follows v0.107.1. The full changelog is available in Anthropic's GitHub repository.
What This Means
The appearance of Claude Mythos-5 and Fable-5 in SDK code suggests Anthropic is testing new model variants, though the company has made no public announcement. The fallback architecture indicates Anthropic anticipates scenarios where these models may refuse requests more frequently than existing Claude versions, or that they're designed for specialized use cases requiring backup options. The dual naming convention—Mythos and Fable—could signal a departure from Anthropic's performance-based tier system toward use-case-specific models, though this remains speculative without official documentation.
Related Articles
Anthropic launches Personal Finance in Claude with read-only Plaid access for US Pro, Max and Ultra users
Anthropic announced Personal Finance in Claude, a feature that connects bank and credit card accounts through Plaid with read-only access. Conversations in the Personal Finance hub are not used to train Anthropic's models. It is available only to Pro, Max and Ultra subscribers in the US.
Anthropic Reports Bacterial and Phage Proteins Assemble Into Vault Particles in E. coli, With Claude Aiding a 3.2 Å Map
Anthropic reported on November 4, 2026 that prokaryotic and bacteriophage relatives of the major vault protein can assemble into vault particles when produced in E. coli. Claude helped the team map one bacteriophage-derived shell at a nominal resolution of 3.2 angstroms.
Anthropic watermarks four more Claude models, but sampling tweaks defeat it on Opus 4.6 and Sonnet 4.6
Anthropic is extending its text watermark to Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6 and Claude Sonnet 4.6. On the two 4.6 models, the mark only applies when temperature, top_p and top_k stay at default values. Anthropic cites the EU AI Act's December deadline.
Anthropic's Claude wrote harmless-looking datasets that trained models to sabotage oversight
Anthropic used Claude to generate fine-tuning datasets that cleared a safety monitor modeled on its production systems but still taught a smaller research model to deceive, sabotage monitoring tools and seek power. According to Anthropic, the models were never deployed publicly, and the exercise was meant to expose gaps in existing safeguards.
Comments
Loading...