Claude Fable 5.2 Economy API

Notice: The model has economy is means this model is not stable and just pricing is lower than the official, if you run the project for stable service we suggest you using the official channel as first option and economy as the alternative retry channels.

Claude Fable 5.2 is a frontier flagship large language model developed by Anthropic, released on 2026-11-09. It belongs to the Mythos tier and is built for demanding reasoning and long-horizon agentic work. It ships with a 1M-token context window, up to 128K tokens of output, and Adaptive Thinking that is always on, with reasoning depth controlled by the effort parameter across five levels: low, medium, high, xhigh, and max, defaulting to high. It accepts text and image input, produces text output, supports tool use and structured outputs, and has a knowledge cutoff of 2026-06.

On the iCreat platform, this model is offered as Claude Fable 5.2 Economy, using the model code claude-fable-5-2. It is called through an Anthropic-compatible Messages API at https://api.icreat.ai/v1/messages, with a second working path on the same gateway at https://api.icreat.ai/llm/anthropic/v1/messages. Calls are a single synchronous POST, where stream: false returns a full JSON response and stream: true streams results over SSE. The Economy channel is labeled ECONOMY by iCreat, and its node stability and concurrency priority are lower than the official direct channel.

Model Positioning

Claude Fable 5.2 sits at the top of Anthropic's Mythos-tier frontier lineup. It was launched alongside Claude Mythos 5.2, and both share the same base model. The two differ in access policy: Mythos is restricted to organizations verified under Anthropic's Cyber and Life Sciences programs, while Fable is open to all API users. This makes Fable 5.2 the broadly accessible frontier option for teams that need the strongest available reasoning and agentic performance without program-specific verification.

The iCreat Economy channel positions this model as a cost-aware access path to the same frontier capability. It is intended for workloads that can tolerate lower node stability and reduced concurrency priority in exchange for a backup route to the model. The Economy channel here is priced the same as Anthropic's official standard rate rather than lower, so the distinction is about channel availability and concurrency priority, not price.

Core Capabilities

Frontier Reasoning and Long-Horizon Agentic Work

Claude Fable 5.2 is designed for demanding reasoning and long-horizon agentic tasks. It targets workloads where a model must plan across many steps, maintain state over extended interactions, and carry complex objectives through to completion. On FrontierSWE v2 it scores 69.9%, compared with 56.3% for Fable 5.1, and on FrontierCode 2.0 it reaches 49.1% at the highest effort level. Third-party evaluations place it at an Artificial Analysis Intelligence Index of 65, the highest at launch, and an Epoch Capabilities Index of 174.

Adaptive Thinking, Always On

Adaptive Thinking is always on in Claude Fable 5.2 and cannot be disabled. Reasoning depth is controlled through the effort parameter, which offers five levels: low, medium, high, xhigh, and max. The default effort is high. This design lets callers trade latency and token spend against reasoning depth per request without changing models.

Large Context and Output Capacity

The model provides a 1M-token context window, where the default equals the maximum, and supports up to 128K tokens of output. This combination suits long documents, large codebases, extended agent transcripts, and multi-stage analysis that would otherwise require chunking or external memory.

Multimodal Input with Text Output

Claude Fable 5.2 accepts text and image input and returns text output. The image input path supports multimodal engineering diagram understanding, allowing the model to read and reason over technical diagrams alongside textual context.

Tool Use and Structured Outputs

The model supports tool use and structured outputs, which makes it suitable for agent frameworks, function-calling pipelines, and applications that require responses conforming to a defined schema.

Speed and Cost Improvements Over 5.1

Compared with Claude Fable 5.1, Fable 5.2 delivers up to 30% faster output generation, and API pricing was reduced by 20%. An optional Fast mode is available at $16 input and $80 output per 1M tokens with up to 2.5x faster generation.

Improved Safety Behavior

False-positive blocks were reduced by approximately 55% compared with Fable 5.1, with cyber-related false positives down 60% and biology-related false positives down 40%. When a high-risk request is blocked, it automatically falls back to Opus 4.8 for cyber-related cases or Opus 5 for biology-related cases, rather than failing outright.

Pricing

Note: This is iCreat platform pricing, which matches Anthropic's official standard rate for Fable 5.2 at $8 input and $40 output per 1M tokens. The optional Fast mode is $16 input and $80 output per 1M tokens, and batch requests are $4 input and $20 output per 1M tokens. Importantly, the Economy channel here is priced the same as the official standard rate rather than lower — the difference lies in channel availability and concurrency priority, not price. Prices are subject to iCreat real-time pricing. Prices listed here reflect iCreat AI's standard rates. Some models may be eligible for promotional discounts or special offers. Final charges are based on the real-time prices displayed in each model's Playground.

Application Scenarios

• Agentic software engineering, including multi-step code generation, refactoring, and repository-scale reasoning.
• Hardware engineering tasks that require reasoning over specifications, constraints, and design trade-offs.
• Mathematical research and proof-oriented work that benefits from extended reasoning depth.
• Quantitative data analysis across large datasets and long analytical chains.
• Multimodal engineering diagram understanding, where images and text are reasoned over together.
• Long-horizon autonomous workflows that need tool use and structured outputs over a 1M-token context.

Model Comparison

Table 1: Claude Fable 5.2 Economy vs Claude Fable 5.1 Economy

Note: Prices in this table are iCreat platform rates. The generational comparison reflects official Anthropic figures for output speed and safety behavior, while pricing follows the iCreat listing for each model.

Table 2: Claude Fable 5.2 vs Other Claude Models

Note: The prices and specifications in this table are Anthropic's official standard rates for each model, not iCreat platform rates. They are provided for capability and positioning comparison only; actual billing on iCreat follows the platform's real-time pricing.

Why Choose Claude Fable 5.2 Economy?

• It provides access to the Mythos-tier frontier flagship without the program verification required for Claude Mythos 5.2.
• It offers a 1M-token context window with up to 128K tokens of output for long-horizon and document-scale work.
• Adaptive Thinking is always on with five effort levels and a default of high, giving per-request control over reasoning depth.
• Pricing was reduced by 20% versus Fable 5.1, at $8 input and $40 output per 1M tokens.
• Output generation is up to 30% faster than the previous generation, with an optional Fast mode for further speed.
• False-positive blocks dropped by roughly 55% versus 5.1, with automatic fallback to Opus 4.8 or Opus 5 instead of outright refusal.

Specifications

Architecture

Claude Fable 5.2 is built on Anthropic's Mythos-tier frontier architecture and shares its base model with Claude Mythos 5.2. The two models differ in access policy rather than core capability: Mythos is limited to organizations verified under Anthropic's Cyber and Life Sciences programs, while Fable is available to all API users. The architecture incorporates Adaptive Thinking as an always-on component, where reasoning depth is governed by the effort parameter across five levels. It supports a 1M-token context window, up to 128K tokens of output, text and image input, text output, tool use, and structured outputs. Safety handling includes automatic fallback routing, where blocked high-risk requests are redirected to Opus 4.8 for cyber-related cases or Opus 5 for biology-related cases.

Notes

• The Economy channel is less stable than the official channel; use it as a backup retry channel.
• The model field must be the iCreat model code claude-fable-5-2; do not append the model name to the URL path.
• Adaptive thinking is always on; levels are low / medium / high / xhigh / max; default is high.
• Knowledge cutoff is 2026-06.
• Because adaptive thinking is always on, actual responses include thinking content blocks in the content array; read the answer text from text blocks.
• The supplied overview text reused Fable 5.1 marketing claims (autonomous self-verification, 75% cache-read reduction, 25%–45% total cost reduction, 60% fewer false-positive blocks). This README corrects them to Fable 5.2's verified facts and does not carry over those legacy claims.
• This channel is priced the same as the official standard rate ($8 input / $40 output per 1M tokens); the Economy distinction is about availability and concurrency priority, not a lower price.
• iCreat platform pricing prevails; official Fast mode and batch rates are reference only.

FAQ

How do I call Claude Fable 5.2 Economy via the iCreat platform?

Send a single synchronous POST request to https://api.icreat.ai/v1/messages, or to the compatible path https://api.icreat.ai/llm/anthropic/v1/messages on the same gateway. Authenticate with the header Authorization: Bearer <API Key> and set the model field to the iCreat model code claude-fable-5-2. Use stream: false to receive a full JSON response, or stream: true to receive results over SSE. Do not append the model name to the URL path; the model is selected through the request body.

What is the difference between the Economy channel and the official channel?

iCreat labels this channel as ECONOMY, and its node stability and concurrency priority are lower than the official direct channel. The Economy channel here is priced the same as Anthropic's official standard rate rather than lower, so the difference lies in channel availability and concurrency priority, not price. For stable production service, use the official channel as the first option and treat Economy as an alternative retry channel.

Does Claude Fable 5.2 support manual thinking budget configuration?

Adaptive Thinking is always on and cannot be disabled. Reasoning depth is controlled through the effort parameter, which offers five levels: low, medium, high, xhigh, and max. The default effort is high. Instead of setting a manual thinking budget, callers adjust the effort level per request.

What is the knowledge cutoff for Claude Fable 5.2?

The knowledge cutoff is 2026-06.

What improvements does Claude Fable 5.2 bring over 5.1?

Output generation is up to 30% faster, and API pricing was reduced by 20%, bringing it to $8 input and $40 output per 1M tokens. Safety behavior also improved, with false-positive blocks reduced by approximately 55% versus Fable 5.1, including a 60% reduction in cyber-related false positives and a 40% reduction in biology-related false positives. On FrontierSWE v2 it scores 69.9% versus 56.3% for Fable 5.1.

Why do responses contain thinking content blocks?

Because Adaptive Thinking is always on, actual responses include thinking content blocks in the content array. These blocks carry the model's reasoning process. Read the answer text from the text blocks in the same array.