Models/Anthropic/Claude Fable 5
Anthropic

Claude Fable 5

anthropic/claude-fable-5
Live on DIT
Get API Key

Anthropic's most capable widely released model for demanding reasoning, autonomous knowledge work, coding, vision, and long-horizon agents.

ModalitiesText + Image + File Text
DIT in / out$6 / $30per 1M tokens
Context1Mreviewed context window
Max output128Ktokens per response
ReleasedJun 9, 2026anthropic compatible
Capabilities
Always-on adaptive thinking
Tool calling
Structured outputs
Vision
Long context

Overview

Recommended use cases and a precise technical profile, generated from the same reviewed model record used throughout this page.

Long-running autonomy

Ambiguous, multistep work that can take hours or days, with persistent focus and self-correction.

Recommended fit
Frontier engineering

Codebase-wide migrations, complex debugging, production implementation, and sophisticated multi-agent workflows.

Recommended fit
Advanced knowledge work

High-stakes research, financial analysis, scientific reasoning, visual inspection, and enterprise deliverables.

Recommended fit
Technical profileReviewed model metadata
Model IDclaude-fable-5ProviderAnthropic
ProtocolANTHROPIC compatibleDIT availability
5 active routes
InputText, Image, FileOutputText
Context window1,000,000 tokensMaximum output128,000 tokens
ReleasedJun 9, 2026Knowledge cutoffNot published
Reasoning controls
Official identifiers and selectable reasoning effort
Aliases
No alias published
Reasoning effort
low
medium
high
xhigh
max
Default: high
Official tool support
First-party tools documented for this model
Web search
Web fetch
Code execution
Memory
Programmatic tool calling
Context editing
Compaction
Vision
8 documented tools
Usage rules
Official token thresholds and billing multipliers
Cache write1.25×
Multipliers apply to the official base rate; DIT market rates are shown in Pricing.
Supported API endpointsOfficial model API surface
EndpointPathStatus
Messages/v1/messages
Supported
Message Batches/v1/messages/batches
Supported
Not supportedExplicit limitations in the official model documentation
Thinking disabled
Raw chain of thought
Audio input
Audio output
Video input
Video output
Manual extended thinking
Assistant prefilling

Providers

Compare DIT market routing with the official upstream reference across price, speed, and availability. Supplier identities remain private.

Market coverage5 active routes
Selection strategyPrice + health + automatic failover
Runtime window30 days
ProviderInput /MOutput /MCache read /MLatencyThroughputUptime
DIT Market
40% off
Aggregated supply · 5 active routes
$10$6$50$30$1$0.63523.4 msP50 TTFT47.6 tpsP50 output80.54%30-day success
Anthropic official
Reference
Direct upstream · reviewed list price
$10$50$1Not publishedNot publishedOperationalChecked Aug 31, 2026

Prices are USD per 1M tokens. DIT performance and uptime appear only after 50 eligible requests; supplier identities and customer traffic remain private.

Pricing

Compare DIT market rates directly with reviewed official prices so savings remain visible and auditable.

Current price comparison
Effective DIT rate compared with the reviewed official list price.
Effective
USD / 1M tokens
Up to 40% lower
Direct rate comparison · reviewed standard prices per 1M tokens
Price composition
DIT effective rate mix
$36.6combined rate
Cached inputInputOutput
Relative rate composition, not estimated invoice share
Price comparisonReviewed source ↗
UsageOfficialDIT marketSavingBilling basis
Input tokens$10$6~40%Per 1M tokens
Output tokens$50$30~40%Per 1M tokens
Cached input$1$0.6~40%Per 1M cached tokens

Performance

Latency, throughput, activity, and request success use privacy-safe DIT runtime aggregates only when the eligibility threshold is met.

Throughput
47.6 tok/s
P50 across eligible DIT traffic
Latency
3523.4 ms TTFT
Lower is better
Runtime sample
27,118
30-day privacy-safe window
All locations
P50 / P95
30-day eligible DIT traffic · privacy-safe aggregates
Latency distribution
Time to first token and end-to-end
Lower is better · aggregated eligible requests only
Traffic activity
Privacy-safe normalized request index
Relative shape only; customer traffic totals remain private
Request success
Eligible DIT request completion rate
80.54%
System probes and user fallback traffic excluded
DIT runtime distribution30-day eligible traffic
MetricP50P95Reading
Time to first byte3523.4 ms14554.7 msLower is better
End-to-end latency5907.5 msLower is better
Output throughput47.6 tpsHigher is better

Uptime

A concise service-health view with scope and source stated next to the number.

80.54%30-day eligible request success
~27K eligible requestsSystem probes and user fallback traffic excluded

Benchmarks

Reviewed evaluation results with configuration notes and source links kept alongside every score.

Benchmark distribution
Exact reviewed scores across published evaluations.
0–100 scale where the source publishes an index or percentage
Reviewed evaluations2 results
BenchmarkCategoryScoreConfigurationSource
GPQA Diamond
Evaluation
81.2%AutoExacto benchmark result for OpenRouter auto-routing; independently maintained evaluation.Source ↗
TAU-Bench
Evaluation
81.4%AutoExacto benchmark result for OpenRouter auto-routing; independently maintained evaluation.Source ↗

Quick Start

A copy-ready request for the compatible DIT endpoint using this model's reviewed identifier.

cURLanthropic compatible
curl https://api.dit.ai/v1/messages \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-fable-5",
    "max_tokens": 256,
    "messages": [{"role":"user","content":"Explain dynamic model routing."}]
  }'

Sources

Field-level provenance keeps pricing, limits, capabilities, and benchmarks independently reviewable.

Reviewed sources4 references · reviewed Aug 31, 2026
SourceTypeCoversChecked
Anthropic models overview
Official
Pricing, Context, Modalities, Maximum output, Reasoning effortAug 31, 2026
Introducing Claude Fable 5 and Claude Mythos 5
Official
Technical profile, Capabilities, Endpoint, Official tool support, Not supported, Usage rulesAug 31, 2026
Claude Fable 5 and Claude Mythos 5 launch
Official
Released, Technical profile, BenchmarksAug 31, 2026
OpenRouter Claude Fable 5
OpenRouter
BenchmarksAug 31, 2026

Official sources take priority. Third-party metadata is retained only when its field and review date are shown.

Models like Claude Fable 5

Reviewed models ranked by capability overlap, protocol compatibility, and context similarity.

Frequently asked questions

Answers use the same reviewed facts and pricing fields presented in the tables above.

What is Claude Fable 5?+

Anthropic's most capable widely released model for demanding reasoning, autonomous knowledge work, coding, vision, and long-horizon agents.

How much does Claude Fable 5 cost?+

The reviewed official rate is $10 per 1M input tokens and $50 per 1M output tokens. DIT pricing is shown only when a live DIT route exists.

What is the context length of Claude Fable 5?+

Claude Fable 5 supports a reviewed context window of 1,000,000 tokens and up to 128,000 output tokens.

What capabilities does Claude Fable 5 support?+

Reviewed capabilities include Always-on adaptive thinking, Tool calling, Structured outputs, Vision, Long context. Consult the linked official documentation for endpoint-specific limits.

What inputs and outputs does Claude Fable 5 support?+

Reviewed input modalities: Text, Image, File. Reviewed output modalities: Text.

Which API endpoints and reasoning levels does Claude Fable 5 support?+

Claude Fable 5 is documented for Messages, Message Batches. Reasoning effort can be set to low, medium, high, xhigh, max, with high as the documented default.

Is Claude Fable 5 available on DIT?+

Claude Fable 5 currently has 5 active routes. DIT selects eligible supply by price and health, with automatic failover when another route is available.

When was Claude Fable 5 released?+

Claude Fable 5 was released on Jun 9, 2026.