Claude Fable 5
Claude Fable 5 is a flagship-tier model from Anthropic in its Fable family, accepting text and image input with a 1,000,000-token context window.
API Pricing
Cheapest input on Deep Infra · cheapest output on Deep Infra| Provider | Input / 1M | Output / 1M | Cached / 1M |
|---|---|---|---|
| $5.00 | $25.00 | $0.500 | |
| $10.00 | $50.00 | - | |
| $10.00 | $50.00 | - | |
| $10.00 | $50.00 | $1.00 | |
| $10.00 | $50.00 | $1.00 | |
| $10.00 | $50.00 | $1.00 |
Prices updated daily. Last check: Oct 23, 2026
Compare API pricing for every Claude model →Claude Fable 5 pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Reasoning & Knowledge
- GPQA Diamond92.6%
- Humanity's Last Exam55.5%
Coding
- SciCode61.0%
Agentic & Tool Use
- Terminal-Bench Hard62.9%
- Terminal-Bench v2.184.6%
- τ²-bench98.5%
- τ-bench Banking38.1%
Instruction & Long Context
- IFBench63.5%
- Long-Context Reasoning82.3%
Benchmarks measured Oct 2026. Scores are independent evaluations, not vendor-reported.
Model Details
General
- Creator
- Anthropic
- Family
- Fable
- Tier
- Flagship
- Context Window
- 1.0M
- Knowledge Cutoff
- Jan 2026
- Modalities
- Text, Image
Capabilities
- Tool Calling
- Yes
- Open Source
- No
- Aliases
- claude-fable-5, Claude Fable 5, anthropic.claude-fable-5
Strengths & Limitations
Strengths
- 1,000,000-token context window, large enough for full document collections or large codebases in a single request
- Accepts image input alongside text, so screenshots, charts, and scanned pages can be processed without a separate vision model
- Tool calling is supported, enabling function-calling and agentic orchestration patterns
- Flagship tier within Anthropic's Fable family, the configuration aimed at the family's hardest tasks
- Knowledge cutoff of January 2026 per Anthropic's published metadata
- Third-party throughput and latency measurements are available from Artificial Analysis, giving an independent performance reference point
Limitations
- Measured time to first token of roughly 60 seconds (Artificial Analysis) makes it unsuitable for real-time chat or autocomplete UIs
- Output throughput of about 65.7 tokens/sec is moderate, so long generations take noticeable wall-clock time
- Flagship-tier models are generally the most expensive option in a family; lighter Fable variants may be more economical for simple tasks
- Filling the full 1M-token context has cost and latency consequences even when the window technically allows it
- We do not track published benchmark scores (reasoning, coding, math) for this model, so capability comparisons rely on vendor claims
Key Features
About Claude Fable 5
Common Use Cases
Claude Fable 5 fits workloads where context size and input breadth matter more than response latency: analyzing large document sets or contract archives in one pass, reasoning over an entire codebase, summarizing long meeting or research corpora, and multi-step agent runs that accumulate long tool-call histories. Image input extends this to mixed-media jobs such as reviewing slide decks, invoices, technical diagrams, or UI screenshots alongside written instructions. Because the measured time to first token is around a minute, it is better matched to batch pipelines, scheduled analysis jobs, and background agents than to interactive assistants; for user-facing chat or streaming-dependent UX, a lower-latency model — including smaller variants if available — is usually the more practical choice.
Frequently Asked Questions
How much does Claude Fable 5 cost?
Pricing varies by provider and by pricing type — input versus output tokens, batch versus real-time, and any caching discounts a provider offers. Check the pricing table on this page for current rates across the providers we track.
What is Claude Fable 5 best used for?
Long-context analysis and agentic workflows. The 1,000,000-token window suits full-repository code review, large document collections, and extended agent histories, while image input covers mixed text-and-visual material such as slides, diagrams, and screenshots. It is best used where total job completion matters more than instant first response.
Is Claude Fable 5 fast enough for interactive applications?
Generally no. Artificial Analysis measured a time to first token of roughly 60 seconds and output throughput of about 65.7 tokens per second. That first-token latency is far too high for chat interfaces or code autocomplete, though it is acceptable for batch and background processing.
Can Claude Fable 5 process images?
Yes. Its confirmed modalities are text and image, so images can be supplied as input alongside text prompts. We do not have information on image generation or other output modalities.
Does Claude Fable 5 support tool calling?
Yes, tool calling is supported, which allows it to be used in function-calling setups and agent frameworks. Specific implementation details, such as parallel tool execution or structured output modes, depend on the provider's API surface.
How does Claude Fable 5 relate to other Fable models?
It sits at the flagship tier of Anthropic's Fable family, meaning it is the configuration positioned for the family's most demanding tasks rather than for high-volume, cost-sensitive work. If Anthropic offers smaller Fable variants, those would typically trade capability for lower cost and latency.