Fam, peep this.
Anthropic just shipped Claude Fable 5.1 (and Mythos 5.1 for trusted access only).
Same base price as Fable 5. $10 in / $50 out per million tokens.
What changed is the agent bill. Cache reads got cut 75%, down to $0.25 per million tokens.
Anthropic says that lands around 25% cheaper on typical work, and up to about 45% on highly agentic runs.
That's us. Long Claude Code sessions. Big context. Tools firing over and over.
Fable 5.1 is also stronger on long-running coding. Terminal-Bench 4.0 55.8% vs 42% on Fable 5. They also say Claude Code should see about 60% fewer false-positive cyber safeguard interruptions.
If you're still on Fable 5 for heavy loops, try `claude-fable-5-1` and lean on caching hard.
Drop what you see on your first long session. Bill and quality. I want receipts.