LATEST MODEL
View in the Herd

Claude Fable 5.2

Anthropic ⚔️ The Knight Released November 2026

Anthropic's most intelligent generally available model, for demanding reasoning and long-horizon agentic work: agent sessions that run for hours, multistep deep research, and analysis carried through to a finished document, spreadsheet, or deck

Claude Fable 5.2

Anthropic • November 2026

Latest

Training Data

Up to June 2026

Claude Fable 5.2

November 2026

Parameters

Not disclosed

Training Method

Advanced Constitutional AI with always-on adaptive thinking (high effort by default), cyber and biology safeguards, and anti-distillation protections

Context Window

1,000,000 tokens

Knowledge Cutoff

June 2026

Key Features

Up to 30% Faster Output than Fable 5.1 • ~55% Fewer Unnecessary Safeguard Blocks • 1M Context / 128K Output

Capabilities

Agentic Coding: Frontier (69.9% FrontierSWE v2)

Reasoning: Frontier (74.2% Humanity's Last Exam with tools)

Computer Use: Frontier (89.3% OSWorld 2.1)

What's New in This Version

A step above Opus 5.5 on almost every benchmark Anthropic reported: 49.1% on FrontierCode 2.0 (Fable 5.1 32.4%, Opus 5.5 45.5%), 41.3% on the Remote Labor Index (Fable 5.1 17.9%), and more than double Fable 5.1 on RoboDojo Sim (64.9% vs 27.6%); up to 30% faster output, about 55% fewer unnecessary blocks, and cheaper than Fable 5.1 at $8/$40 per 1M tokens (from $10/$50)

Anthropic's most intelligent generally available model, for demanding reasoning and long-horizon agentic work: agent sessions that run for hours, multistep deep research, and analysis carried through to a finished document, spreadsheet, or deck

What's New in This Version

A step above Opus 5.5 on almost every benchmark Anthropic reported: 49.1% on FrontierCode 2.0 (Fable 5.1 32.4%, Opus 5.5 45.5%), 41.3% on the Remote Labor Index (Fable 5.1 17.9%), and more than double Fable 5.1 on RoboDojo Sim (64.9% vs 27.6%); up to 30% faster output, about 55% fewer unnecessary blocks, and cheaper than Fable 5.1 at $8/$40 per 1M tokens (from $10/$50)

Technical Specifications

Parameters Not disclosed
Context Window 1,000,000 tokens
Training Method Advanced Constitutional AI with always-on adaptive thinking (high effort by default), cyber and biology safeguards, and anti-distillation protections
Knowledge Cutoff June 2026
Training Data Up to June 2026

Key Features

Up to 30% Faster Output than Fable 5.1 ~55% Fewer Unnecessary Safeguard Blocks 1M Context / 128K Output

Capabilities

Agentic Coding: Frontier (69.9% FrontierSWE v2)
Reasoning: Frontier (74.2% Humanity's Last Exam with tools)
Computer Use: Frontier (89.3% OSWorld 2.1)
Theme
Language
Support
© funclosure 2025