Anthropic · API

Claude Fable 5 API

Claude Fable 5 is Anthropic's most capable widely released model for demanding reasoning and long-horizon agents, with 1M context and always-on adaptive thinking.

Use 1M context with one OmniaKey API key.

Get your API key Read the docs

Model ID: claude-fable-5
Context: 1M
Provider: Anthropic
Input · USD / 1M tokens: $2.9 USD; Official price $10 USD
Output · USD / 1M tokens: $14.5 USD; Official price $50 USD
Savings: 71%
Last verified: 2026-08-13

What this model is for

Anthropic positions Claude Fable 5 as its most capable widely released model, built for demanding reasoning and long-horizon agentic work. Its published specification combines a 1M-token context window, up to 128K output tokens, vision, tool use, and adaptive thinking that is always enabled.

Use the claude-fable-5 model ID when a task must investigate, act, and verify across a session larger than a single sitting. Fable is a specialist rather than a routine default: compare accepted outcomes against Opus 5 and Sonnet 5 before paying its higher input and output rates at scale.

Use Fable 5 for unusually difficult, long-running agent work when a serious Opus attempt or your own evaluation shows that the higher token rate earns a better completed result.

Best-fit coding workloads

Unresolved root-cause work

Revisit difficult incidents or cross-system bugs after a careful Opus attempt still misses the cause.

Long-horizon agent workflows

Sustain investigation, implementation, tests, and verification across unusually long autonomous tasks.

Multimodal technical analysis

Combine large document sets, code, screenshots, diagrams, and structured tools in one evidence-heavy workflow.

Technical specification

Catalog facts and integration details verified for this route.

Model ID: claude-fable-5
Context: 1M
Inputs and tools: Vision · Tool use · Adaptive thinking
API endpoint: /v1/chat/completions
Cache hit / 1M: $0.29 USD
Coding clients: Claude Code · Cursor · Cline · Aider

Limits to plan around

Fable 5 has the highest listed Claude input and output rates and Anthropic labels its comparative latency as slower. Opus 5 or Sonnet 5 can be more economical when they pass the same acceptance test.
Adaptive thinking is always on. The API does not support thinking type disabled; use the effort control to change reasoning depth, and do not expect raw chain-of-thought output.
Fable 5 safety classifiers can decline a request. Anthropic's Messages API reports a refusal as HTTP 200 with stop_reason refusal, so integrations must test response handling and an approved fallback path.

Compare nearby models

Compare workload fit and current listed token rates before choosing a route.

Model	Best for	Context	Input	Output
Claude Fable 5This page claude-fable-5	The hardest long-running reasoning and agent tasks where evaluations justify the lineup's highest rate.	1M	$2.9 USD Official price $10 USD	$14.5 USD Official price $50 USD
Claude Opus 5 claude-opus-5	Complex coding, architecture, and high-cost-of-error work at half Fable's listed token rates.	1M	$1.2 USD Official price $5 USD	$6 USD Official price $25 USD
Claude Sonnet 5 claude-sonnet-5	Daily implementation, review, and agent loops where speed and cost per iteration are the priority.	1M	$0.48 USD Official price $2 USD	$2.4 USD Official price $10 USD

SDK quickstarts

Use one base URL and model ID from cURL, Python, or TypeScript.

Streaming chat completion

curl https://api.omniakey.com/v1/chat/completions \
  -H "Authorization: Bearer $OMNIAKEY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-fable-5",
    "messages": [
      {"role": "user", "content": "Review this pull request"}
    ],
    "stream": true
  }'

Estimate token cost

Enter expected input and output tokens to compare current listed rates.

Input tokensOutput tokens

OmniaKey estimate: $0.58
Official-rate estimate: $2.00
Estimated savings: $1.42

Estimate covers input and output tokens only. Cache usage, retries, tools, and client-side token accounting can change the final charge.

Related coding guides

Deeper selection, setup, and troubleshooting guidance for this model family.

Claude Fable 5 vs GPT-5.6 Sol for Coding

Test Fable 5 first for the hardest long-horizon agent work; start with GPT-5.6 Sol when Codex, Responses, and the OpenAI tool stack define the workflow.

Read guide

Claude Fable 5 Pricing and API Access

Anthropic lists Fable 5 at $10 input, $1 cache read, and $50 output per million tokens. OmniaKey currently lists $2, $0.20, and $10 respectively.

Read guide

Claude Code Pricing

Choose a subscription for a recurring allowance and hosted features, or API billing when you need metered usage, explicit model costs, and project-level spend control.

Read guide

Sources and verification

Model facts and links were checked on 2026-08-13. Provider documentation remains the authority for upstream behavior.

Frequently asked questions

How much does Claude Fable 5 cost on OmniaKey?

OmniaKey charges $2.9 USD per 1M input tokens and $14.5 USD per 1M output tokens, in USD. Official prices are $10 USD and $50 USD; you save 71%.

What is the context window for Claude Fable 5?

The listed context window is 1M. Actual usable limits can also depend on the client and upstream request format.

When should I choose Claude Fable 5 instead of Opus 5?

Choose Fable 5 when the task is exceptionally long, ambiguous, and tool-heavy, or when a controlled evaluation shows it completes work that Opus 5 does not. Repository size alone is not enough; use Opus or Sonnet when they meet the same acceptance criteria at a lower rate.

Can adaptive thinking be disabled on Claude Fable 5?

No. Anthropic documents adaptive thinking as always on for Fable 5, and thinking type disabled is unsupported. Use the effort parameter or Claude Code's effort control to trade reasoning depth, latency, and token use instead.

How should an integration handle a Claude Fable 5 refusal?

Anthropic's Messages API can return HTTP 200 with stop_reason refusal rather than an HTTP error. Test how your selected OmniaKey protocol surface exposes that result, avoid retry loops, and define an approved model fallback before production use.