Claude Fable 5.1 vs Claude Opus 5.5: when to pay 2.5x

Standard API rates in US dollars, as published by each maker on September 26, 2026. Batch and priority tiers, taxes, and subscription plans are not included. Claude Fable 5.1: The full 1M context window is billed at standard rates. Claude Opus 5.5: The full 1M context window is billed at standard rates. Fast mode, a research preview on the Claude API, costs $8 input and $40 output per million tokens.

How much more does Claude Fable 5.1 cost than Opus 5.5?

Claude Fable 5.1 charges $10 per million input tokens and $50 per million output tokens. Claude Opus 5.5 asks $4 and $20 for the same tokens. Cache writes keep the same 2.5x ratio: $12.50 and $20 for 5-minute and 1-hour writes on Fable 5.1, against $5 and $8 on Opus 5.5. Any work that skips the cache pays the full multiple, so the large one-off review costs $2.00 against $0.80 and the output-heavy generation $4.30 against $1.72.

Cache hits are where the two models come closest. Anthropic bills a hit at 0.025x the input price on Fable 5.1 and 0.05x on Opus 5.5, so a cached million tokens costs $0.25 on one and $0.20 on the other, a 1.3x gap. On Fable 5.1 that discount is steep enough that reading 2M tokens from the cache in the example session costs $0.50, just 5% of the session total.

Because reads are such a small slice, the cheap hit narrows the session gap only a little: $10.50 against $4.40, or 2.4x. The money goes to cache writes, which cost $6.50 on Fable 5.1 and make up 62% of its session. Of the $6.10 difference per session, $3.90 comes from writes and $1.50 from output. At 110 sessions a month, that is $1,155.00 against $484.00 at API rates.

When does Anthropic suggest Fable 5.1 over Opus 5.5?

Anthropic's guidance describes a default and an escalation. It calls Opus 5.5 its recommended starting model for most work and says, "It performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5." Fable 5.1 is Anthropic's most capable generally available model, aimed at demanding reasoning and long-horizon agentic coding, and Anthropic suggests it when Opus-tier results fall short.

That points to a working pattern rather than a single pick: run Opus 5.5 by default and move a task to Fable 5.1 when the result isn't good enough. In Claude Code, /model fable switches to it. Each model keeps its own cache in Claude Code, though, so the first request after a switch writes the whole context again at Fable 5.1's $12.50 or $20 per million.

The strength claims overlap. Anthropic lists long-running agentic coding and multistep research for Fable 5.1, and agentic coding, computer use, and codebase-wide migrations and audits for Opus 5.5. This page can't check either claim, but it can price them, and on identical token counts the premium for Fable 5.1 is 2.5x on every uncached token.

Effort, data retention, and fast mode

Both models accept 1M tokens of context at standard rates, write up to 128K tokens of output, and use Anthropic's newer tokenizer, which counts about 30% more tokens than earlier Claude models for the same text. A prompt is the same size on either, so the rate card alone sets the gap for input.

Output is less predictable, because the effort defaults differ. Fable 5.1 starts at high and Opus 5.5 at medium, and Anthropic notes that changing effort on Opus 5.5 keeps the prompt cache. Output already makes up 24% of the Fable 5.1 session in the example, and a higher effort setting tends to raise how much each request writes, at $50 per million.

Two facts can settle the question outright. Fable 5.1 requires 30-day data retention, so it is not available under zero data retention. And for latency-sensitive work, Opus 5.5 offers fast mode, a research preview on the Claude API at $8 input and $40 output per million, still below Fable 5.1's standard rates. You'll find both in Claude Code, Cursor, OpenRouter, OpenCode, and GitHub Copilot.

Anthropic

Claude caches a prompt prefix up to a breakpoint. One top-level cache_control field places the breakpoint automatically and moves it as the conversation grows, or you can mark up to 4 blocks yourself. Claude Code manages caching for you.

A 5-minute cache write costs 1.25x the input price and a 1-hour write costs 2x. A cache hit costs 0.1x input on most models, 0.05x on Claude Opus 5.5, and 0.025x on Claude Fable 5.1 and Claude Mythos 5.1. Every hit restarts the cache lifetime at no charge.

Anthropic's rule of thumb: a 5-minute write pays for itself after one cache read, and a 1-hour write after two.

In Claude Code, the main conversation uses the 1-hour cache on a Claude subscription and the 5-minute cache with an API key. Each model has its own cache, so switching models starts over.

Source: Anthropic: Prompt caching

Sources

1. Anthropic: Pricing
2. Anthropic docs: Claude Fable 5.1
3. Anthropic: Claude Fable 5.1 and Claude Mythos 5.1
4. Claude Code docs: Model configuration
5. Cursor docs: Claude Fable 5.1
6. OpenRouter: Claude Fable 5.1
7. OpenCode docs: Zen
8. GitHub Docs: Supported AI models in Copilot
9. Anthropic docs: Claude Opus 5.5
10. Anthropic: Claude Opus 5.5
11. Anthropic docs: Fast mode
12. Cursor docs: Claude Opus 5.5
13. OpenRouter: Claude Opus 5.5
14. Anthropic: Prompt caching