treyleo16/ fable-5-1

61 1 week ago

For your toughest challenges.

cloud 
ollama run treyleo16/fable-5-1

Models

View all →

1 model

fable-5-1:latest

415kB · - context window · Text · 1 week ago

Readme

Claude Fable 5.1

Claude Fable 5.1 is Anthropic’s flagship Mythos-class reasoning engine engineered for long-horizon agentic workflows, codebase-wide software engineering, and complex, multi-hour knowledge tasks. Positioned as the general-availability counterpart to Claude Mythos 5.1, Fable 5.1 delivers frontier-tier problem solving while introducing substantial API cost optimizations for cached context and tool-heavy agent loops.

Model Overview

Fable 5.1 is designed for multi-step tasks that run autonomously over extended periods—spanning complex file edits, continuous code reviews, and deep analytical research without cutting corners or hallucinating premature success.

  • Developer: Anthropic
  • Model Class: Mythos-Class Frontier Reasoning Engine
  • Release Date: September 1, 2026
  • Architecture: Autoregressive Transformer with Always-On Adaptive Thinking & Vision
  • Primary Target Use Cases: Autonomous multi-hour agent execution, multi-repository codebase refactoring, complex financial/legal document synthesis, and scientific hypothesis testing.

Core Capabilities & Architectural Improvements

Sustained Multi-Hour Agentic Work

  • Honest Problem Solving: Unlike models that take shortcuts or force failing tests to pass, Fable 5.1 explicitly identifies root causes and reports blockers when stuck rather than falsely claiming success.
  • Long-Horizon Persistence: Built to run unattended across multi-hour task loops, managing error recovery, background verification, and progress updates autonomously.

Cost-Optimized Context Caching

  • 75% Reduced Cache Read Cost: Drops prompt cache read pricing down to $0.25 per MTok. This cuts overall API costs by ~25% on standard workloads and up to ~45% on tool-heavy, agentic workflows.

Refined Safeguards & Enterprise Privacy

  • Defensive Cybersecurity Access: Updated safety filters allow Fable 5.1 to detect and patch software vulnerabilities natively, reducing false-positive safety flags by 60% compared to previous generations.
  • Enterprise Frontier Safeguards (EFS): Supports customer-controlled cloud infrastructure deployment options, granting zero data retention (ZDR) privacy guarantees.

Technical Specifications

Parameter Specification
API Model ID claude-fable-5-1
Context Window 1,000,000 tokens (1M)
Max Output Tokens 128,000 tokens (128k)
Input Modalities Text, Code, Images, Structured PDFs
Output Modalities Text, Code, Structured JSON
Reasoning Engine Adaptive Thinking (Always-On, High effort default)
Input / Output Pricing \(10.00 / MTok Input \| \)50.00 / MTok Output ($0.25 Cache Read)

Fable 5.1 vs. Mythos 5.1 Relationship

Model Target Audience & Access Safety & Safeguard Configuration
Claude Fable 5.1 General Availability: Available to Pro/Max/Team/Enterprise users and standard Cloud APIs. Standard production safeguards; defensive cybersecurity allowed, exploit generation blocked.
Claude Mythos 5.1 Trusted Access Only: Project Glasswing, vetted defense, & life sciences partners. Extended capabilities for specialized biological binder design and research workflows.