If you use Claude Code or Claude with MCP, the notable part is that the most capable model is now cheaper than its predecessor.
What's new / How it works
- Models: Claude Fable 5.2 (API name
claude-fable-5-2) and Claude Mythos 5.2. Anthropic states they are the same underlying model; Mythos 5.2 is limited to organizations in the Cyber and Life Sciences Verification Programs, while Fable carries cybersecurity and biology safeguards. - Pricing: $8 per million input tokens and $40 per million output tokens, 20% below Fable 5.1 ($10 / $50). An optional fast mode costs double.
- Speed and quality: Anthropic reports improvements in coding, reasoning and vision and output up to 30% faster than the 5.1 versions. These are the vendor's own results and not independently verified.
Why it matters
Long agentic runs – generating components, rewriting tokens, checking UI against screenshots – burn many tokens. Lower prices and better vision affect both cost and the quality of visual checks. For everyday design work the biggest model may not be necessary; smaller models (Sonnet, Haiku 5.5) may suffice for less.
How to start
- In Claude Code, switch to Fable 5.2 (model settings or
/model). - Give it a task an earlier model struggled with, such as comparing an implementation to a design screenshot.
- Compare the result and token usage with a cheaper model to decide where it pays off.