Fable 5
The most powerful AI model ever made publicly available.
Built on Anthropic's Mythos architecture.
80.3% SWE-Bench Pro · 1M context window · 128K output tokens · Adaptive thinking always on
Fable 5 & Mythos 5
Same underlying weights. Same Mythos-class intelligence. Different safety profiles for different use cases. Fable 5 is the public twin — Mythos 5 is restricted to vetted cyber defence partners.
Claude Fable 5
Public release- Available to all Pro, Max, Team & Enterprise users
- Safety classifiers route <5% of queries to Opus 4.8
- High-risk topics (cyber, bio, chem) fall back to Opus
- Full API access: claude-fable-5
- Amazon Bedrock & direct API
- GitHub Copilot integration
Claude Mythos 5
Restricted access- Restricted to vetted cyber defenders & infrastructure operators
- No safety classifiers — full unrestricted Mythos capabilities
- Deployed via Project Glasswing (US gov collaboration)
- Same model ID: claude-mythos-5
- Same pricing as Fable 5
- Application-only access through Anthropic
State-of-the-art on
nearly every benchmark.
The longer and more complex the task, the larger Fable 5's lead. Exceptional performance in software engineering, knowledge work, vision, scientific research, and finance.
Software Engineering
First model to break 80% on SWE-Bench Pro — 11 points ahead of the next frontier model. Handles complex multi-file refactors, debugging, and full-stack development autonomously.
Vision & Documents
Reads diagrams, charts, tables nested in PDFs, scanned documents, and screenshots. Opens up research and document-heavy work in finance, legal, analytics, and architecture.
Analytics & Finance
First model to break 90% on complex, long-running analytical tasks — a 10-point jump over Opus. The strongest finance-first model ever tested on general finance and reasoning.
Scientific Research
Deep understanding of scientific papers, experimental design, and cross-domain synthesis. Handles literature review, hypothesis generation, and methodology critique at expert level.
Knowledge Work
The longer and more complex the task, the larger Fable 5's lead. Strategic memos, policy drafts, competitive analysis, and comprehensive reporting — all at a level exceeding Opus.
Tool Use & Agents
Near-perfect tool orchestration with multi-step planning, parallel execution, and adaptive error recovery. Powers the most sophisticated AI agent workflows on any platform.
Numbers don't lie.
Fable 5 compared against Claude Opus 4.8, GPT-5.5, and Gemini 3.1 Pro across key benchmarks.
Benchmark data from Anthropic's official release. Scores represent verified external evaluations.
What you can build
with Mythos-class AI.
Autonomous coding agents
Fable 5 writes, debugs, and deploys production code across entire repositories. Multi-file refactors that took hours now complete in minutes with 80.3% accuracy on SWE-Bench Pro.
Financial analysis & modelling
Process earnings reports, build DCF models, and generate investment memos with 91.2% accuracy on complex analytical tasks. The strongest finance model ever benchmarked.
Legal document review
Analyse contracts, identify risk clauses, and compare regulatory frameworks across jurisdictions. 93.1% accuracy on document understanding with full vision capabilities.
Scientific research synthesis
Review literature, critique methodologies, and generate hypotheses across domains. Cross-references papers and datasets at a level previously requiring specialist researchers.
Enterprise knowledge work
Draft strategy decks, competitive analyses, and policy documents. The longer the task, the larger Fable 5's advantage — designed for complex, multi-hour knowledge work.
Production AI agents
Orchestrate multi-step workflows with near-perfect tool use (96.1%). PURIST deploys Fable 5-powered agents for lead qualification, customer onboarding, and operations automation.
Mythos power,
responsible deployment.
Fable 5 is built from the same weights as Mythos 5 but includes safety classifiers that trigger in less than 5% of sessions. High-risk queries are automatically routed to Claude Opus 4.8 — the model stays powerful while remaining safe for general use.
Project Glasswing
The collaboration between Anthropic and the U.S. government that governs how Mythos-class models are deployed. Fable 5 is the public-safe output of this programme.
How Fable 5 compares.
API pricing per million tokens. Fable 5 offers Mythos-class performance at a lower cost per token than Opus.
| Model | Input | Output | Context |
|---|---|---|---|
| Claude Fable 5 Mythos-class | $10 | $50 | 1M tokens |
| Claude Opus 4.6 Opus-class | $15 | $75 | 200K tokens |
| Claude Sonnet 4.6 Sonnet-class | $3 | $15 | 200K tokens |
| Claude Haiku 4.5 Haiku-class | $0.80 | $4 | 200K tokens |
Availability
Fable 5 & Mythos 5 released
Available on Pro, Max, Team, and Enterprise plans at no extra cost
Free access period ends
Fable 5 moves to usage credits. API pricing: $10/$50 per million tokens
API access via Anthropic & AWS Bedrock
Model ID: claude-fable-5 — available on Amazon Bedrock and direct API
Fable 5, deployed
for your business.
PURIST integrates Fable 5 into production n8n workflows for lead qualification, document processing, financial analysis, and autonomous operations. Mythos-class intelligence, deployed in days — not months.