Anthropic releases Claude Fable 5 and Mythos 5 with major gains in coding and science

· Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Life Sciences & Biology · Depth: Intermediate, medium

Summary

Anthropic has released its fifth-generation AI models, Claude Fable 5 and Claude Mythos 5, on June 9, 2026, demonstrating significant advancements in various domains. Fable 5, designed for general use, achieves top scores in programming, image processing, and complex data analysis benchmarks. It scored 80.3% on SWE-Bench Pro and 29.3% on FrontierCode, outperforming Claude Opus 4.8, GPT 5.5, and Gemini 3.1 Pro. Mythos 5, initially available to select partners, excels in specialized areas like drug design, speeding up processes by 10x, and autonomous genomics research, where it outperformed a Science-published model. It also scored 78% on ExploitBench for cybersecurity. Both models are priced at \$10 per million input tokens and \$50 per million output tokens, nearly doubling the cost of Claude Opus 4.8. Fable 5 incorporates safety guardrails, routing high-risk requests to Opus 4.8, while Mythos 5 has fewer restrictions for its specialized applications.

Key takeaway

For AI Engineers evaluating advanced models for high-stakes applications, you should consider Claude Fable 5 for its superior performance in coding and complex analysis, noting its \$10 per million input token cost. If your work involves sensitive areas like cybersecurity or drug design, explore Mythos 5 through Project Glasswing, understanding its specialized access and fewer safety restrictions. Be prepared for higher operational costs and potential initial exclusion from standard subscription plans.

Key insights

Anthropic's Fable 5 and Mythos 5 models set new benchmarks in coding, science, and specialized applications, albeit at a higher cost.

Principles

Method

Fable 5 employs AI classifiers to identify and redirect high-risk requests in cybersecurity, biology/chemistry, and distillation to Claude Opus 4.8, limiting their effectiveness.

In practice

Topics

Best for: CTO, VP of Engineering/Data, Machine Learning Engineer, AI Scientist, AI Engineer, Director of AI/ML

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.