Claude Fable 5.1 is Anthropic’s premium model for difficult, long-running knowledge work. It is built for complex coding agents, research, document creation, and vision-heavy workflows—not as the default model for every API request.
The practical decision is simple: Fable 5.1 offers Anthropic’s highest-end capability at $10 per million input tokens and $50 per million output tokens, while Claude Opus 5 costs half as much. Start with Opus 5 for normal production work. Move selected tasks to Fable 5.1 when stronger reasoning or sustained autonomous execution measurably improves the result.
Claude Fable 5.1 specifications
| Specification | Claude Fable 5.1 |
|---|---|
| Claude API model ID | claude-fable-5-1 |
| Release date | September 1, 2026 |
| Context window | 1 million tokens |
| Maximum output | 128,000 tokens |
| Input | Text and images |
| Output | Text |
| Standard API price | $10/M input, $50/M output |
| Batch API price | $5/M input, $25/M output |
| Prompt-cache read | $0.25/M tokens |
| License/access | Proprietary, broadly available |
Anthropic lists a June 2026 knowledge cutoff. The model supports tool use and adaptive thinking, which is always enabled. Developers can control how much effort it uses, but should expect higher latency than lighter Claude models.
Where Fable 5.1 is available
Direct Anthropic API requests use claude-fable-5-1. Anthropic also documents availability through Amazon Bedrock, Google Cloud, Microsoft Foundry, and Claude Platform on AWS. Provider billing, regions, model names, and feature support can differ, so a marketplace listing is not interchangeable with direct Anthropic access.
OpenRouter lists the model as anthropic/claude-fable-5.1 at the same $10/$50 standard token rates. That is a routed provider option, not a different Anthropic model.
Anthropic’s Batch API applies a 50% discount, but support is not universal across every marketplace. At launch, Anthropic documents batch processing on the Claude API and Claude Platform on AWS—not on Bedrock, Google Cloud, or Microsoft Foundry. Do not append a fictional “batch model” suffix; batch is a request mode for the same model.
Minimal API example
import anthropic
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-fable-5-1",
max_tokens=4096,
messages=[{"role": "user", "content": "Review this migration plan and identify failure modes."}],
)
print(message.content[0].text)For production agents, add explicit timeouts, retries, spend limits, and a fallback model. Anthropic does not automatically replace Fable 5.1 with another model when a request fails. Our AI API failure guide covers retry and fallback design, while the AI testing and evaluation hub explains how to evaluate model changes before rollout.
Best developer use cases
Long-running coding agents
Fable 5.1 is aimed at difficult repository-scale work: multi-file migrations, architecture analysis, debugging across services, and tasks where an agent must maintain a plan over many tool calls. That does not mean every coding prompt benefits. Route simple edits and routine tests to a cheaper model, then escalate hard tasks.
Research and large document sets
The 1M context window can hold substantial codebases or document collections. Context size is a capacity limit, not a quality guarantee. Retrieval, source selection, and evaluation still matter.
Vision and document workflows
The model accepts images alongside text, which supports screenshot analysis, diagrams, scanned documents, and visual debugging. It produces text rather than generated images.
High-stakes security work
Anthropic applies additional safeguards for advanced cyber and biological capabilities. Access to a powerful model does not remove permission, review, or sandboxing requirements. For agent credentials and tool permissions, use the controls in our AI application security guide.
Fable 5.1 versus Opus 5
| Decision factor | Fable 5.1 | Opus 5 |
|---|---|---|
| Input/output price | $10/$50 | $5/$25 |
| Context | 1M | 1M |
| Best starting point | Hardest sustained work | Most complex production work |
| Cost strategy | Escalation tier | Default premium tier |
Anthropic itself recommends starting most workloads with Opus 5 and using Fable 5.1 when the task justifies the premium. That is more useful than treating vendor benchmark wins as a universal ranking. Read our updated Opus 5 vs Fable 5.1 decision guide for a workload-by-workload comparison.
How Mythos 5.1 relates
Claude Mythos 5.1 uses the same underlying model, specifications, capabilities, and pricing as Fable 5.1, but its access model is different. Anthropic positions Mythos for vetted US organizations through Project Glasswing and other trusted-access arrangements. It is not a generally available public API tier.
Do not use claude-mythos-5-1 as though it were a normal public model ID, and do not infer Mythos access from ordinary Fable availability.
Limitations to plan for
- Cost: output is $50/M tokens before batch discounts.
- Latency: adaptive thinking and hard tasks can be slower.
- Provider differences: batch support and model identifiers vary.
- Vendor benchmarks: Anthropic’s published results are useful evidence, not independent reproduction.
- Data handling: review retention and regional-processing requirements before sending sensitive data. Anthropic also offers US-only inference at a 1.1× token-price multiplier.
Verdict
Claude Fable 5.1 is a selective escalation model, not a blanket replacement. Use it when a complex agent, large research task, or difficult multimodal workflow produces enough additional value to justify twice the Opus 5 token cost. For the rest, keep Opus 5 or a cheaper model as the default and evaluate Fable on a representative dataset before changing production routing.