Anthropic dropped two models today.
Most people are focused on the wrong one.
What launched:
Fable 5 — the public-facing Mythos-class model. Available on Pro, Max, Team, and Enterprise right now. Stronger than anything Anthropic has released publicly.
Better at long-context, agentic coding, vision, and knowledge work. The longer and more complex the task, the bigger the gap between Fable and previous models.
Mythos 5 — same architecture, fewer restrictions. Going out through Project Glasswing to a small group of security providers and government institutions only. You're not getting access to this one. Neither am I.
The catch nobody's leading with:
Fable 5 is free on your subscription until June 22. On June 23 it moves to usage credits.
Pricing is $10/million input tokens and $50/million output tokens — double what you're paying now for Opus.
You have 12 days to hammer it before it costs real money.
What's actually different for developers:
The safeguard fallback is interesting architecture. Sensitive queries — cybersecurity, biology, chemistry — silently fall back to Opus 4.8.
Anthropic says this triggers in under 5% of sessions. If you're building in those domains, test your workflows now because your production behaviour may shift depending on which model actually responds.
For agentic work the jump is real. I build automation workflows with n8n and Claude and the longer the chain, the more Fable pulls ahead. Multi-step reasoning, document processing, complex conditionals — noticeably sharper.
What I'm actually testing this week:
- Multi-agent loops on my n8n stack
- Long-context RAG retrieval quality vs Opus 4.7
- Vision tasks on mixed document types
The bigger picture:
Mythos-class capability is now public. With guardrails, yes — but the ceiling just moved up significantly. The question isn't whether Mythos ever ships publicly. It's how fast this capability tier becomes the default baseline.
Use the next 12 days well.
What are you testing Fable 5 on first? Drop it in the comments.