Claude Fable 5 price: when e make sense to use am

Claude Fable 5 cost na how much?

Claude Fable 5 cost $10 for each million input tokens and $50 for each million output tokens for Claude API. Na Anthropic publish these list rates, based on the pricing page wey dem read on 10 October 2026. The API model ID na claude-fable-5. E become generally available on 9 June 2026 for Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud and Microsoft Foundry. Anthropic now list am as legacy model. The successor, Claude Fable 5.1 (claude-fable-5-1), release on 1 September 2026 with the same input and output prices. Cache reads for am cost $0.25 for each million tokens, na one quarter of Fable 5 rate. Fable 5 still dey available, and dem no go retire am before 9 June 2027.

[
  {
    "label": "Claude Fable 5",
    "input": "10",
    "output": "50",
    "cache_read": "1",
    "batch_input": "5",
    "batch_output": "25"
  },
  {
    "label": "Claude Opus 5",
    "input": "5",
    "output": "25",
    "cache_read": "0.50",
    "batch_input": "2.50",
    "batch_output": "12.50"
  },
  {
    "label": "Claude Sonnet 5",
    "input": "2",
    "output": "10",
    "cache_read": "0.20",
    "batch_input": "1",
    "batch_output": "5"
  },
  {
    "label": "Claude Sonnet 4.6",
    "input": "3",
    "output": "15",
    "cache_read": "0.30",
    "batch_input": "1.50",
    "batch_output": "7.50"
  },
  {
    "label": "Claude Haiku 4.5",
    "input": "1",
    "output": "5",
    "cache_read": "0.10",
    "batch_input": "0.50",
    "batch_output": "2.50"
  }
]

Make you see am as a ladder. Fable 5 list price na two times Claude Opus 5, five times the current Claude Sonnet 5 rate, and ten times Claude Haiku 4.5. The ratio na the same for both sides of the bill, because every model for that chart price output at exactly five times its own input rate. So if you sabi the input and output token split for one job, you fit convert one price to any other by multiplication.

Sonnet 5 list price na $2 and $10 for each million tokens. Anthropic first sell that price as introductory rate until 31 August 2026, then make am the standard price and cancel the planned increase to $3 and $15. So the five times gap between Fable 5 and Sonnet 5 still hold, and budget for autumn fit use the same Sonnet numbers.

The discounts, and the one multiplier that goes the other way

Prompt caching na the biggest lever. Cache read dey cost one-tenth of the base input price, wey be $1 per million tokens for Fable 5. To write to cache dey cost pass ordinary input token: 1.25 times base price for the five minute cache, and 2 times base price for the one hour cache. Five minute cache don pay for itself after one read, while one hour cache pay after two reads. Na this calculation make e important to work out where prompt caching breaks even before you turn am on everywhere.

Batch API dey remove 50% from both sides, so batched Fable 5 work dey cost $5 and $25 per million tokens. Batch requests dey asynchronous, so this one only help work wey no need answer immediately. The two discounts fit stack, so cached, batched job cheap for both input and output.

The 1M token context window dey included for standard rates on Claude 4.6 and later models, with one exception: Claude Haiku 5.5. Fable 5 no get long context surcharge, so 900k token request dey bill for the same per-token rate as 9k token request. Haiku 5.5, released on 7 October 2026, dey price based on prompt length instead. Once prompt pass 100,000 tokens, input rate go rise from $0.10 to $0.50 per million tokens, and output rate go rise from $0.50 to $2.50.

The multiplier wey dey increase the bill na data residency. If you set inference_geo: "us" to keep inference inside the United States, e go apply 1.1 times multiplier to every token category, including cache reads and cache writes. Leave the default global routing unless contract require another setting.

One billing rule dey apply specifically to this model family. Fable 5 ships safety classifiers wey fit decline request. When this happen, Messages API go return stop_reason: "refusal" as successful HTTP 200 response instead of error. Whether you go pay for refusal wey arrive before any output depend on its stop_details.category. As of 10 October 2026, refusals for the bio, frontier_llm and reasoning_extraction categories dey billed like any other request, using the rates for the model wey run am. Refusal before any output for any other category, or with null category, no dey billed. Anthropic talk say this list fit change, so check the category table for its refusals documentation instead of assuming say refusal dey free. If you retry the same prompt on another Claude model, fallback credit go refund the prompt cache cost of the switch, so you no go pay to warm the same cache twice.

Which plans include Claude Fable 5?

As of 10 October 2026, Anthropic help centre article about Fable models talk say Fable 5 and Fable 5.1 dey available for all paid plans (Pro, Max, Team and Enterprise). E also talk say Fable 5 no dey available for Free plan, and plan comparison table for Anthropic pricing page show No for Fable row under Free. Pro na the cheapest seat for that list, so wetin Pro cost and where e usage limits stop you set the minimum cost for any subscription route to Fable. If Fable access na the main reason you wan upgrade, first work through whether e worth am to pay for Pro seat based on your own usage, because small Fable allowance no be strong enough reason for subscription by itself. For developers, e generally dey available for Claude API and through the cloud marketplaces we list above. Anthropic pricing page no list free tier for that route, and as of 10 October 2026, neither the page nor help centre describe any signup credit. Read wetin the free route to Claude API really cover before you rely on one. Paid seat na now the exception. Since 7 October 2026, Max subscribers dey get $100 (Max 5x) or $200 (Max 20x) every month as Claude Platform credit, and Team plans fit get up to $500 every month shared across seats. The credit work with any Claude model wey dey available, including Fable 5. Free, Pro and Enterprise plans no qualify, and unused credit go expire at the end of each billing cycle.

Model wey dey available for plan no mean say e dey included without limit. Subscription no dey meter exactly like API either, because Pro seat no get any published token quota and e count usage through messages inside rolling window instead. So nothing for plan convert cleanly to the per million figures above. Anthropic pricing page plan comparison table put Fable behind part of weekly usage limits for some seats and behind usage credits for others, and this arrangement change more than once during July 2026. Anthropic own statement about redeploying Fable 5 include the model for up to 50% of weekly usage limits through 7 July 2026 on Pro, Max, Team and selected Enterprise plans, with usage credits after that date. Reports for the rest of July describe more extensions and different treatment between seat types, and help centre now record say the promotion end on 19 July 2026. As of 10 October 2026, Max seats, premium Team seats and premium seats for seat based Enterprise plans fit use up to 50% of their weekly usage limits on Fable models without extra charge. That share come from the normal weekly limit and e no add to am. Pro, standard Team and standard seat based Enterprise seats run Fable with pay as you go usage credits from the beginning.

So this page no go print a per-plan limit in messages or tokens, and even that 50% share na only wetin help centre talk on the day wey person check am. No figure wey dem publish during July last for two weeks, and stale number here go worse than no number. Open the row labelled Fable for plan comparison table on the day you decide, and treat any number for news article as out of date. If you take seat because of Fable access and the allowance come be smaller than wetin the table make you expect, going back down before the next renewal leave the month wey you don already pay for untouched.

The thing wey stable na API rate. For API, Fable 5 bills according to the prices for the chart above. For subscription seat, after you pass the included share, you either pay with usage credits or stop. Usage based Enterprise plans make this structure clear, because Enterprise seat fee buy access while the tokens still meter at API rates on top of am. Premium seats for seat based Enterprise plans na the exception, because Fable dey included inside the 50% share wey we describe above. That make the price list the figure to plan with either way. Na the same conclusion you reach when you compare API billing with subscription for any other Claude model. If you never choose tier yet, start with which Claude plan match your usage. If you go use that seat as your everyday assistant, instead of using am only to get Fable, e make sense to check how Claude plans compare in price with ChatGPT before you commit to one.

Wey your bill dey pass wetin price ratio suggest

Two documented mechanisms dey make Fable 5 cost pass wetin straight price comparison imply.

The first one na tokenizer. Fable 5 dey use the tokenizer wey Claude Opus 4.7 introduce. Compared with models wey release before Opus 4.7, the same text dey produce roughly 30% more tokens, and the exact increase depend on the content. Haiku 4.5 come before that tokenizer, so the ten times gap for the price list dey closer to thirteen times when you feed both models the same text block. Sonnet 5 dey use the newer tokenizer, so Fable against Sonnet na fair per token comparison. Fable against Haiku no be fair comparison.

The second one na thinking. Adaptive thinking dey always on for Fable 5, and thinking: {"type": "disabled"} no dey supported. Dem dey bill thinking tokens as output tokens, with the full output rate. This model no dey return the raw chain of thought, and thinking.display default to "omitted", meaning say short answer wey user fit see fit carry plenty billed output count. Anthropic documentation talk am clearly: the billed output token count no match the visible token count for the response.

Read the breakdown instead make you guess. The usage.output_tokens_details.thinking_tokens field report how many billed output tokens go for reasoning:

{
  "usage": {
    "input_tokens": 25,
    "output_tokens": 348,
    "output_tokens_details": {
      "thinking_tokens": 312
    }
  }
}

For that example, wey come from Anthropic thinking documentation, 312 of the 348 billed output tokens na reasoning wey nobody ever see. When you stream, this breakdown dey arrive only for the final message_delta event, so client wey stop to read after the last text chunk no go ever record am.

The control na effort, wey you set for output_config.effort, with levels low, medium, high (the default), xhigh and max.

{
  "model": "claude-fable-5",
  "max_tokens": 32000,
  "output_config": { "effort": "medium" },
  "messages": [{ "role": "user", "content": "..." }]
}

Anthropic guidance for this model na make you start with high, move up to xhigh only for work wey capability matter pass, and step down to medium or low for routine work. The reason na say lower effort for Fable 5 often beat xhigh for older models. Two traps dey here. If you change effort between requests, your prompt cache go become invalid, because the resolved effort value dey render inside the prompt. So choose one level for each workload and keep am constant. Fable 5.1 add per message effort change for beta, and e dey preserve the cache, but Fable 5 no support am. And max_tokens na hard limit for total output, including thinking and response text, so cap wey you size for answer without thinking fit truncate the response. If you see stop_reason: "max_tokens", either raise the cap or lower the effort.

Cost per completed task, no be cost per token

[
  {
    "label": "Short answer (5k in, 1k out)",
    "fable_5": "0.10",
    "opus_5": "0.05",
    "sonnet_5": "0.02",
    "haiku_4_5": "0.01"
  },
  {
    "label": "Coding session (300k in, 40k out)",
    "fable_5": "5.00",
    "opus_5": "2.50",
    "sonnet_5": "1.00",
    "haiku_4_5": "0.50"
  },
  {
    "label": "Long agent run (2M in, 400k out)",
    "fable_5": "40.00",
    "opus_5": "20.00",
    "sonnet_5": "8.00",
    "haiku_4_5": "4.00"
  }
]

Those 3 rows na arithmetic, dem no be benchmark. Dem don publish the rates. The token volumes na assumptions wey you suppose replace with figures from your own usage data. Coding session wey resemble the middle row go cost $5.00 for Fable 5, compared with $1.00 for Sonnet 5. The long run go cost $40.00 compared with $8.00. The Haiku column for that last row na arithmetic only: Haiku 4.5 get 200k token context window, so e no fit hold that job at all.

Cost per token no be the correct unit for decision. Cost per completed task na cost per attempt multiplied by the number of attempts. With today's rates, one Fable 5 attempt cost the same as five Sonnet 5 attempts, so retry count alone almost never justify the upgrade. Sonnet go need fail more than four times out of five before the cheaper path lose on tokens.

The thing wey justify am na everything wey token bill leave out. The gap between the two long runs for the chart small pass one hour of engineer time for most markets. If the expensive model save one hour of review, or prevent one bad migration wey you go later need undo, e don already pay for itself through bill wey no show the saving. Do the comparison with money per accepted result, and include your own time for the total. Know wetin one million tokens really fit buy go make that estimate far less abstract.

Jobs wey Claude Fable 5 fit justify im price

Long-running agent work wey wrong step fit cost plenty. Fable 5 dey made for work wey fit take hours: e get 1M token context window, up to 128k output tokens for each request, and an xhigh effort level wey target tasks wey dey run pass thirty minutes with token budgets wey reach millions. Compare the run cost with the cleanup work after agent take wrong branch for step 40, and nobody notice am until step 300.

Migration or audit wey you go run once across big codebase. One-off work no get enough volume to spread the cost, so the per-token premium go enter one invoice, and the whole job fit stay inside one context window. If e fit run asynchronously, Batch API go cut the cost by half to $25 for each million output tokens.

Task wey cheaper model don fail twice already. The two failed attempts plus your debugging time cost pass one Fable attempt at the volumes for the chart above. Escalation na the cheaper branch, and na wetin model ladder dey for. If you still dey choose between tiers generally, the capability comparison between the Claude model tiers dey answer different question from the one wey this page dey answer.

Jobs three wey Sonnet 5 or Haiku 4.5 be the honest answer

High volume classification and extraction. Dem dey grade these jobs based on throughput and unit cost, and the quality ceiling low reach enough say cheap model fit reach am. If task cost ten times the price wey Haiku 4.5 use complete correctly, Fable 5 no dey buy anything wey you fit measure. Haiku 4.5 don become legacy model now, and dem no go retire am before 15 October 2026. Anthropic docs point to Claude Haiku 5.5 as the model to migrate to, so read wetin Haiku 5.5 cost and where e long prompt pricing dey start before you build new pipeline on the older one.

Interactive work where latency na the product. Anthropic current model table rate Fable 5.1, wey be Fable 5 successor for the same price, as slower; na the slowest rating for the lineup. You no fit turn off Thinking for either Fable model. Chat box or editor completion dey feel worse for the more capable model, because users notice the wait before dem notice the quality.

Anything wey depend on events after January 2026. Fable 5 reliable knowledge cutoff na January 2026. Opus 5 own na May 2026. The more expensive model na the less current one, so for recency questions you pay double for older knowledge unless you give am search or fetch tools. Fable 5.1 move the Fable cutoff go June 2026.

One constraint dey outside cost completely. Dem designate Fable 5 as Covered Model, so e require 30 day data retention and e no dey available under zero data retention unless Anthropic expressly authorize am. If your agreement require zero retention, Fable 5 harder to access, but dem never rule am out. Organisation wey get ZDR arrangement fit turn on 30 day retention for one workspace and run Fable there, while the other workspaces keep zero retention. Anthropic 1 September 2026 Enterprise Frontier Safeguards announcement also talk say eligible customers go receive ZDR for Fable 5 and Fable 5.1 until those safeguards ready. Without that authorization or workspace wey get retention turned on, the API go reject Fable 5 requests, and no discount fit change that.

fable-method repos show wetin, and wetin dem no show

Practitioners don publish repos wey distil how Fable 5 approaches dey work into skills wey cheaper model fit follow. The fable-method repo describe one thinking skill, one orchestration loop, and one adversarial verifier wey dey run every claimed check again instead of just reading the agent own report. Its README talk am directly: "This na community distillation, e no be Anthropic artifact."

Treat am exactly like that. E show how people dey drive the model, but e no be documentation of how the model dey work. Anthropic no validate anything inside am, and several almost identical forks dey with different wording. Na wetin you go expect from folklore, no be from one spec.

The useful signal for cost decision be say the parts people find worth copying na procedural. Classify the task, name the check wey go prove say e don complete, gather evidence before you decide, then verify by observation instead of reading summary. If you give cheaper model explicit checklist and verification step, e fit close part of the gap. So run that experiment for your own evaluation set before you standardise on the expensive model. The result depend on your tasks, na why other people number no get much value for you.

Check your own numbers before you commit budget

• Read usage for every response, and log output_tokens_details.thinking_tokens separately from the output wey you fit see. The reasoning wey you no see na the line item wey fit surprise people.
• Set effort explicitly instead of accepting the default, and keep am constant inside any conversation wey depend on prompt caching.
• Put your stable prefix behind a cache breakpoint first. Cache read wey cost one tenth of base input price fit save pass wetin most model downgrades go save.
• Move anything wey no need immediate answer go the Batch API.
• Price the alternative honestly. Compare Claude and ChatGPT API pricing for your own workload fit help before you make long-term commitment.

If the workload na agent wey dey run for your own server, the controls wey matter pass dey outside the model choice. Keep agent costs under control on a VPS cover the loop limits and spend caps wey fit stop runaway session, while where Claude Code actually spends tokens explain the numbers wey you go see for your usage dashboard.

FAQ

Claude Fable 5 cost per million tokens na how much?

For Claude API, Claude Fable 5 price na $10 for each million input tokens and $50 for each million output tokens. Anthropic pricing page confirm this on 10 October 2026. Cache read cost na $1 for each million tokens, wey be one tenth of the normal input rate. Batch API reduce both prices by 50%, so asynchronous work cost $5 for each million input tokens and $25 for each million output tokens. If you keep inference inside the United States with the inference_geo parameter, every category get 1.1 times multiplier.

Claude Fable 5 dey inside Claude Pro subscription?

Anthropic help centre list Fable 5 as available for all paid plans (Pro, Max, Team and Enterprise), and e talk say Free plan no support am. Say model dey available for a plan no mean say subscription don include the usage. As of 10 October 2026, Pro users reach Fable through pay as you go usage credits from the beginning. Max seats and premium Team seats fit use up to 50% of their weekly usage limits without extra cost. This arrangement change more than once during July 2026. So, check the row wey carry Fable for the plan comparison table on Anthropic pricing page on the day you wan decide. No rely on figure from an article.

Why my Claude Fable 5 bill high pass wetin visible output show?

Adaptive thinking dey always on for Claude Fable 5. System bill thinking tokens as output tokens, but e no dey return the raw chain of thought. When thinking.display dey use its default value of omitted, you go see empty thinking field, but you still dey pay for every reasoning token behind am. Read usage.output_tokens_details.thinking_tokens for the response to see the breakdown. If you dey stream the response, this information go arrive only for the final message_delta event. Reduce the effort level if reasoning-to-answer ratio high pass wetin the task need.

I suppose use Claude Fable 5 or Claude Opus 5?

Claude Opus 5 cost half the price per token and get newer reliable knowledge cutoff: May 2026 for Opus 5, compared with January 2026 for Fable 5. Start with Opus 5, then escalate if the task fail there, or if wrong answer go cost more than the price difference. Fable 5 make sense for long horizon agent work where one bad branch fit cause hours of cleanup, and for one-off jobs where the premium go appear on one invoice instead of every request forever. Both models don become legacy models. Anthropic docs direct new work to Claude Opus 5.5 at $4 and $20 for each million tokens, or to Claude Fable 5.1 at Fable 5 prices.