Claude Fable 5 Economy
Claude Fable 5 is Anthropic’s officially released Mythos-tier model, designed for autonomous knowledge work, advanced reasoning, and long-horizon coding. It supports text, image, and file inputs and produces text output. It features a 1-million-token context window, up to 128,000 output tokens, and built-in always-on adaptive thinking, tool use, and structured outputs. The model shares the same underlying model family as Claude Mythos 5 but was released publicly with stronger safety safeguards. It is suitable for complex software engineering, multi-step agentic workflows, enterprise research, document analysis, and professional tasks requiring sustained reasoning and high reliability.
Related Models
Claude Fable 5.1 Economy
Claude Fable 5.1 is Anthropic's upgraded flagship frontier model engineered for high-complexity software development, long-horizon agentic workflows, and multi-step knowledge work. Built on Mythos-tier reasoning, Fable 5.1 significantly enhances autonomous self-verification and root-cause troubleshooting while reducing safety false-positive interventions by up to 60%. Powered by an optimized prompt-caching architecture featuring a 75% reduction in cache-read pricing, it cuts total operating costs by 25% to 45% for highly agentic tasks.
Claude Opus 5 Economy
Claude Opus 5 is Anthropic's ultra-flagship frontier model engineered for extreme cognitive density, deep logical reasoning, and expert-level scientific synthesis. Powered by native adaptive thinking and a 1-million-token context window, it resolves complex multi-step mathematical, architectural, and coding challenges with unprecedented accuracy and near-zero hallucination rates. Coupled with rigorous enterprise safety alignment, Claude Opus 5 excels at high-stakes operations, including legal compliance, quantitative financial modeling, advanced biosecurity research, and mission-critical software architecture auditing.
Claude Opus 4.8 Economy
Claude Opus 4.8 is Anthropic’s most powerful officially released Opus model, designed for complex reasoning, long-horizon agentic coding, and highly autonomous professional workflows. It supports text, image, and file inputs and produces text output. It features a 1-million-token context window, up to 128,000 output tokens, and built-in capabilities for adaptive thinking, tool use, and structured outputs. The model excels at advanced coding, browser and computer-use agents, enterprise knowledge work, financial and legal analysis, and multi-step tasks requiring sustained judgment and high reliability.
Claude Sonnet 5 Economy
Sonnet 5 is Anthropic’s most capable Sonnet-class model, delivering frontier-level performance across coding, agentic workflows, and professional tasks. It features adaptive thinking with selectable reasoning levels—**low**, **medium**, **high**, **max**, and **x-high**—along with a 1-million-token context window and support for text, image, and file inputs. Built with an updated tokenizer, Sonnet 5 also incorporates real-time cybersecurity safeguards designed to block high-risk dual-use activities.
Read Me
Claude Fable 5 API
Claude Fable 5 is Anthropic's highest-capability widely released model for ambitious coding, long-running agents, and multi-stage knowledge work. It is designed for tasks that continue across planning, execution, verification, and revision rather than ending after a single response.
With iCreat, you can call claude-fable-5 through our Anthropic-compatible Messages API. The endpoint supports the Anthropic SDK request structure, streamed responses, and content-block handling, giving existing Claude applications a direct path to test the model without operating inference infrastructure.
Model Positioning
Claude Fable 5 sits above Claude Opus 4.8 when the highest available Claude capability matters more than minimum latency. Anthropic positions it for difficult coding and professional work that can run asynchronously for extended periods, including projects that require repeated planning and self-checking.
It is not the default choice for every prompt. Short answers, routine transformations, and latency-sensitive traffic can use a faster model, while Fable 5 handles the smaller share of requests where task completion, autonomy, and reasoning depth justify additional processing.
Key Capabilities
Long-Running Agent Work
Claude Fable 5 is built for agent workflows that preserve a goal across many stages, use intermediate results to revise a plan, and verify work before completion. An application still needs an agent harness, tools, permissions, and stopping rules; the model provides the reasoning and continuity inside that system.
Ambitious Software Engineering
The model targets large migrations, complex implementations, and extended coding sessions where changes must remain consistent across a codebase. It is especially relevant when the workflow includes inspection, implementation, testing, review, and correction rather than isolated code generation.
Always-On Adaptive Thinking
Adaptive Thinking is always enabled at the model level. Claude Fable 5 decides when and how much to reason based on the request, while an effort setting can control reasoning depth on supported endpoints. This makes the model suitable for mixed workflows in which simple steps and difficult decisions occur within the same task.
Long Inputs and Extended Outputs
Anthropic publishes a 1M-token context window and a maximum output of 128K tokens for Claude Fable 5. This supports workflows that combine extensive source material with substantial deliverables, such as a repository-wide implementation plan, a detailed research synthesis, or a long technical artifact.
Visual and Document Reasoning
Claude Fable 5 accepts text and image input through iCreat. It can interpret diagrams, charts, tables, and visual information contained in files and PDFs, supporting workflows that combine written instructions with visual evidence.
Pricing
iCreat bills Claude Fable 5 by input, output, cache-read, and cache-write tokens:
| Token Type | Price |
|---|---|
| Input | $1 per 1M tokens |
| Output | $5 per 1M tokens |
| Cache Read | $0.10 per 1M tokens |
| Cache Write | $1.25 per 1M tokens |
Use Cases
- Complex code migrations: plan and execute coordinated changes across large repositories.
- Long-running agent backends: sustain multi-stage work with checkpoints, tools, and validation supplied by the application.
- Professional research: connect evidence across extensive sources and produce review-ready analysis.
- Document-heavy workflows: reason over dense reports, tables, diagrams, and structured business material at the model level.
- Quality-first routing: send the hardest requests to Fable 5 while keeping routine traffic on a faster model.
These workloads still require application-level evaluation, access controls, tool permissions, and human review when outputs affect production systems or consequential decisions.
Model Comparison
Claude Fable 5 vs Claude Opus 4.8
| Factor | Claude Fable 5 | Claude Opus 4.8 |
|---|---|---|
| Positioning | Highest-capability widely released Claude model | Complex agentic coding and enterprise work |
| Context Window | 1M tokens | 1M tokens |
| Maximum Output | 128K tokens | 128K tokens |
| Reasoning Control | Automatic reasoning-effort adjustment; always on | Automatic reasoning-effort adjustment; optional |
| Best Fit | Long-running, asynchronous, quality-first work | Complex work where lower latency remains important |
Claude Fable 5 vs Kimi K3 and GPT-5.6 Sol
| Factor | Claude Fable 5 | Kimi K3 | GPT-5.6 Sol |
|---|---|---|---|
| Positioning | Long-running autonomous agents | Long-horizon coding and knowledge work | Complex production workflows and tool use |
| Context Window | 1M tokens | 1M tokens | 1.05M tokens |
| Official Input Modalities | Text and image | Text, image, and video | Text and image |
| Reasoning Control | Automatic reasoning-effort adjustment | Manual reasoning-effort adjustment | Manual reasoning-effort adjustment; Pro mode available |
| Best Fit | Multi-stage asynchronous work | Large repositories and visual development | OpenAI tool ecosystem and fine-grained reasoning control |
Why Choose Claude Fable 5?
Choose Claude Fable 5 when the application must carry a difficult objective across many steps and return work that is ready for review rather than a short first draft. Its combination of always-on Adaptive Thinking, a 1M-token context window, 128K maximum output, and Anthropic's focus on long-running agents makes it a quality-first option for demanding coding and knowledge workflows.
For routine prompts or minimum-latency traffic, use a faster model and route only the tasks that benefit from Fable 5's higher capability ceiling.
Specifications
| Category | Description |
|---|---|
| Model Name | Claude Fable 5 |
| Provider | Anthropic |
| Model ID | claude-fable-5 |
| Model Type | Frontier reasoning LLM |
| Context Window | 1M tokens |
| Maximum Output | 128K tokens |
| Official Input Modalities | Text and image |
| Output Modalities | Text |
| Official Reasoning Control | Always-on Adaptive Thinking |
| iCreat Input Modalities | Text and image |
| iCreat API Features | Anthropic-compatible Messages API, text and image input, streaming, Adaptive Thinking, thinking/text content blocks |
| Best Suited For | Long-running agents, ambitious coding, multi-stage knowledge work |
Architecture
Anthropic describes Claude Fable 5 as a Mythos-level model but does not publish its parameter count or underlying network architecture. Its most important visible design choice is always-on Adaptive Thinking: the model determines when and how much reasoning to use instead of relying on a manually assigned thinking-token budget.
Raw chain-of-thought is not returned. Depending on the supported thinking-display configuration, a thinking block contains either a readable summary or an empty field. Applications should treat these blocks as response metadata and rely on the final answer, tools, tests, and external evidence for validation.
Notes
The 1M context window and 128K maximum output are Anthropic's published model specifications. Use the current limits shown on the iCreat model page when sizing production requests.
Adaptive Thinking cannot be disabled at the model level, and reasoning contributes to the request's output budget. Set output limits with both reasoning and the final response in mind, then measure latency and token usage on representative workloads.
Certain sensitive cybersecurity, biology, and chemistry requests may receive stricter safeguards or refusals. Handle refusal responses explicitly, and do not assume that Anthropic's native fallback behavior is also available through the iCreat endpoint.
FAQ
What should a production evaluation measure for Claude Fable 5?
Measure end-to-end task completion, tool errors, retries, human corrections, latency, and total token use. A model can produce an impressive final response while still being costly or unreliable across the full agent trajectory.
How should long-running Fable 5 jobs be checkpointed?
Store the current plan, completed steps, verified tool results, and unresolved decisions in application state. Checkpoint before expensive or irreversible actions so an interrupted job can resume without repeating the entire trajectory.
Should thinking blocks be retained across agent turns?
Yes. Preserve the assistant's returned content blocks when continuing a multi-turn exchange or returning tool results. This maintains reasoning continuity and avoids reconstructing only the visible text portion of the response.
How can an application keep autonomous actions within scope?
Use least-privilege tool permissions, action allowlists, approval gates for consequential operations, time and token limits, and an auditable tool log. The model should not be the only component deciding what it is authorized to do.