Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens, with cached input at $1 and cache writes at $12.50 for the five-minute window or $20 for the hour, read from the Anthropic pricing page on 16 August 2026 and confirmed unchanged on 5 September 2026.
That is exactly twice Claude Opus 5 on every line, and five times Claude Sonnet 5. Mythos 5 carries the same rates on limited availability.
What Claude Fable 5 costs, line by line
Anthropic bills Claude on five lines rather than two, and Fable 5 is no exception. The two cache write columns are what most calculators leave out, and they are the ones that decide whether a long system prompt is cheap or expensive to keep alive.
| Billing line | Fable 5, per million tokens |
|---|---|
| Input | $10.00 |
| Cache write, 5 minute window | $12.50 |
| Cache write, 1 hour window | $20.00 |
| Cached input read | $1.00 |
| Output | $50.00 |
Fable 5 at $10 in and $50 out per million, read 16 August 2026
Two ratios are worth carrying in your head. Output costs five times input, which is the widest gap in the Claude line and means your bill is driven by how much the model writes, not by how much you send it. And a cached read costs a tenth of a fresh one, which is the single largest lever on any workload that repeats context. Our LLM cost calculator prices the same volume across every provider we track.
Fable 5 is priced at exactly double Opus 5 on all five lines. Not on input alone, not on output alone: the whole grid is multiplied by two.
Fable 5 against Opus 5, Opus 4.8 and Sonnet 5
The question behind most searches for this price is not what Fable 5 costs in absolute terms, it is what the step up from Opus buys. Here is the whole ladder on the two lines that matter, read the same day.
| Model | Input / cached / output, per million |
|---|---|
| Claude Fable 5.2, listed 9 November 2026 | $8.00 / $0.20 / $40.00 |
| Claude Mythos 5.2, limited availability | $8.00 / $0.20 / $40.00 |
| Claude Fable 5.1 | $10.00 / $0.25 / $50.00 |
| Claude Mythos 5.1, limited availability | $10.00 / $0.25 / $50.00 |
| Claude Fable 5 | $10.00 / $1.00 / $50.00 |
| Claude Mythos 5, limited availability | $10.00 / $1.00 / $50.00 |
| Claude Opus 5 | $5.00 / $0.50 / $25.00 |
| Claude Opus 4.8 | $5.00 / $0.50 / $25.00 |
| Claude Sonnet 5 | $2.00 / $0.20 / $10.00 |
| Claude Sonnet 4.6 | $3.00 / $0.30 / $15.00 |
| Claude Haiku 4.5 | $1.00 / $0.10 / $5.00 |
Opus 5 and Opus 4.8 are priced identically, so moving from the older Opus to Opus 5 costs nothing, and the newer Opus 5.5 is listed lower still, at $4.00 and $20.00 on the pricing page we read on 29 September 2026. That makes Opus 4.8, and now Opus 5 as well, the wrong thing to stay on for price reasons, and it makes the real decision a two-way one: Opus 5.5 or Fable 5, with Fable 5 at two and a half times Opus 5.5 on input and output. The comparisons below keep Opus 5 as the reference, because that is the grid Fable 5 doubles line for line.
The other line worth noticing is Sonnet. Sonnet 5 is cheaper than the Sonnet 4.6 that preceded it, at $2.00 against $3.00 on input. A new generation that costs less than the one it replaces is rare enough to check twice, and we did: those $2.00 and $10.00 figures were launched as introductory pricing through 31 August 2026, and Anthropic has since confirmed they are the standard price.
What the cache columns change
Cached input is billed at one tenth of fresh input across the Claude line, Fable 5 included. The point releases broke that rule at the top of the range: on Claude Fable 5.1 and Claude Mythos 5.1 a cache read costs a fortieth of fresh input, $0.25 per million against $1.00 on Fable 5. As detailed in the newer Fable 5.1 grid, this makes the newer pair four times cheaper on anything that resends context. Fable 5.2 and Mythos 5.2, listed on 9 November 2026, keep that fortieth on a lower $8 input, so their cache read falls to $0.20. Claude Opus 5.5, listed on the Anthropic pricing page we read on 29 September 2026, adds a third ratio: its cache read costs a twentieth of fresh input, $0.20 per million. On a workload that resends the same system prompt or the same document on every call, that ratio matters more than the model you picked.
Take the same month, two million input tokens and five hundred thousand output. Without cache, Fable 5 costs $45.00. With ninety percent of the input served from cache, the same month costs $28.80. You saved more by caching than you would have saved by dropping from Fable 5 to Opus 5 on output alone.
The cache write is the line most estimates leave out. It is charged once, when the prompt is stored, at $12.50 or $20.00 per million depending on the window you keep it alive for. Whether that is worth paying depends on how often your context changes, which is a question about your application rather than about the grid. We work through the arithmetic in our full Claude API pricing page, cache included.
What a month actually costs
Three shapes of work, priced on the rates above with ninety percent of input served from cache, which is realistic for anything that carries a stable system prompt.
| Monthly workload | Fable 5 / Opus 5 / Sonnet 5 |
|---|---|
| Assistant: 2M in, 0.5M out | $28.80 / $14.40 / $5.76 |
| Same volume, no caching at all | $45.00 / $22.50 / $9.00 |
| Agent: 20M in, 4M out | $238.00 / $119.00 / $47.60 |
The gap is a constant factor of two against Opus 5 and five against Sonnet 5, at every volume, because the whole grid is multiplied rather than the individual lines. That is unusual: on most providers the ratio between tiers moves depending on whether your work is input-heavy or output-heavy. Here it does not, which makes the decision cleaner than it looks. It is not how Gemini or Grok price their own tiers.
One lever halves all of it at once. Requests sent through the Batch API are discounted fifty percent on input and output alike, which puts Fable 5 at $5.00 and $25.00 per million, the standard Opus 5 grid. The trade is that batch work is asynchronous, so it only fits jobs nobody is waiting on. On a model priced at $50 output, that is the largest single reduction available to you, and it stacks with caching.
When Fable 5 is worth twice Opus 5
Since the multiplier is flat, the question is never “is Fable 5 cheap on my workload”. It is “does this task fail on Opus 5”. Two situations where the answer is often yes: work where a single wrong output costs more than the whole month of inference, and work where you would otherwise run several passes on a cheaper model to get one usable result.
That second case is the one people underestimate. Three passes on Opus 5 cost more than one pass on Fable 5 and take three times as long. If your pipeline retries, count the retries before you compare the grids.
The reverse holds too. On high-volume, low-stakes work, the constant factor of five between Fable 5 and Sonnet 5 is the whole difference between a bill you notice and one you do not. Below Haiku the market keeps going, and we mapped where the floor actually sits. Our comparison of what every major model charges per million tokens puts both against the rest of the market.
The same $10 and $50 also buys a faster Opus 5
There is a second way to spend Fable 5 money on the Claude API, and it is easy to miss. Fast mode, in research preview, serves Claude Opus 5 and Opus 4.8 (Anthropic retires the Opus 4.8 option on 30 November 2026) at premium rates, and those rates are $10 per million input and $50 per million output. The exact Fable 5 grid.
So the decision at this price point is not only which model, it is which axis you are buying. The same budget gets you either a different model at standard speed, or the same Opus 5 you already run, answering faster. Fast mode is first party only, it is not available through the Batch API, and the cache multipliers stack on top of it rather than replacing it. If latency is what your product is short of, that is the cheaper experiment of the two, because nothing else in your pipeline has to change.
How you buy it changes the price
Everything above is the rate you pay Anthropic directly. Going through a router adds a layer: OpenRouter charges its own margin on top of the provider grid, and the managed clouds run separate grids again. On a model priced at $50 output, a few percent of markup is worth checking before you route traffic through anything.
Frequently asked questions
How much does Claude Fable 5 cost per million tokens?
$10.00 for input and $50.00 for output, with cached input at $1.00 and cache writes at $12.50 for five minutes or $20.00 for an hour. Read from the Anthropic pricing page on 16 August 2026.
Is Fable 5 more expensive than Opus 5?
Yes, exactly twice as expensive, on every one of the five billing lines. There is no volume or workload shape at which that ratio changes.
What is the difference between Fable 5 and Mythos 5 on price?
None. Mythos 5 carries identical rates, on limited availability. If you can get either, price is not what separates them.
Is there a cheaper way to run Fable 5?
Two, and they stack. The Batch API takes fifty percent off input and output, bringing Fable 5 to $5.00 and $25.00 per million. Prompt caching takes ninety percent off any input you resend. Neither changes the model you get.
Does Claude Enterprise change these rates?
No. An Enterprise seat is $20 per person per month and the tokens are still billed at API rates on top, which we detail in our Claude Enterprise pricing page.
How do I cut a Fable 5 bill without changing model?
Cache the stable part of your prompt and shorten your outputs. Cached reads cost a tenth of fresh input, and output is five times input, so those two levers move the total far more than anything else available to you.
What we verified and what we did not
Verified: all five billing lines for Fable 5, Mythos 5, Opus 5, Opus 4.8, Sonnet 5, Sonnet 4.6 and Haiku 4.5, read from the Anthropic pricing page on 16 August 2026, along with the note confirming that the Sonnet 5 introductory rate is now standard, the fast mode grid for Opus 5, and the fifty percent Batch API discount, with the Opus 4.8 fast mode retirement read in the release notes on 3 November 2026.
Not verified, and therefore not asserted: rates on Bedrock or Vertex AI, which price Claude on their own grids and are covered in our Bedrock page; any benchmark or quality claim, since this page prices models rather than ranking them; and the availability of Mythos 5, which Anthropic marks as limited without publishing what the limit is.
Prices in this category move. Everything here carries the date it was read, and the grid is worth checking at the source before you commit a budget to it.