Anthropic shipped Opus 5.5 on 22 September. If Claude is already in the stack, this is a model swap, not a new tool.
List price versus Opus 5 is 20% down: $5 / $25 to $4 / $20. Cache reads fell 60% from $0.50 to $0.20. Anthropic says 5.5 uses fewer tokens per finished task, writes less jargon, outputs more than 30% faster, and costs about 40% less than Opus 5 on typical agent workloads. Fable 5.1 stays at $10 / $50. Context is 1 million tokens. It is on Anthropic, AWS, Google Cloud, and Azure. Pro, Max, and Team usage limits went up.
There are breaking API changes. Thinking cannot be disabled. Forced tool use returns an error. Thinking blocks stay tied to the model that made them. Read the migration note before you flip production.
For agencies the useful part is long briefs, client-facing drafts, and multi-step coding agents that already live in Claude. You do not get a new workflow. You get a cheaper, faster run of the one you have, if cache is actually hitting.
“Fable-level at 40% off” packs two claims. Quality versus Fable is Anthropic’s eval. The 40% is workload cost, not the $4 / $20 row. Short uncached jobs get about 20%. Long agent loops with cache hits can move more.