Claude Fable 5
AnthropicLarge Language ModelProprietaryAnthropic's most capable widely available model, built for the most demanding reasoning and long-horizon agentic work. State-of-the-art across nearly all AI benchmarks, with exceptional performance in software engineering, scientific research, and vision tasks.
Abilities
Use Cases
Available in Tools
Availability
How to Use
Pros
- State-of-the-art performance on nearly all AI benchmarks — the most capable Claude model publicly available
- 1M token context window with 128K max output — processes entire codebases and book-length documents
- Adaptive thinking always on — automatically applies optimal reasoning depth for every task
- Exceptional software engineering — Stripe reported it compressed months of engineering into days
- Outstanding vision capabilities — extracts data from scientific figures and rebuilds code from screenshots
- Available on all major cloud platforms for enterprise compliance
Cons
- Most expensive widely available Claude model at $10/$50 per MTok
- Slower response times than Sonnet/Haiku due to deeper reasoning
- Safety classifiers may fall back to Opus 4.8 for some cybersecurity and biology queries
- Closed source — no model weights or training details available
What to Use It For
Perfect For
State-of-the-art on SWE-bench and agentic benchmarks — can run autonomously for longer than any previous Claude
Exceptional GPQA Diamond scores and ability to extract insights from complex scientific figures and datasets
Proven in production by Stripe and other enterprises to dramatically accelerate complex development work
Good For
Maintains quality and coherence across very long outputs with adaptive thinking
Excellent at extracting structured data from images, PDFs, and complex visual documents
Not Recommended
At $10/1M input it is too expensive for bulk classification or extraction pipelines
Try instead: Claude Haiku 4.5, Claude Sonnet 5
Slower than Sonnet/Haiku due to deep adaptive reasoning — perceptible delay in interactive conversations
Try instead: Claude Sonnet 5
Do Not Use For
Text and vision LLM only — no image generation capabilities
Try instead: DALL-E 3, FLUX.1
Closed source cloud-only model — completely dependent on Anthropic API
Try instead: Llama 4 Maverick