Media & Culture

Claude Opus 5 matches Fable 5's capabilities with stronger safeguards

Priced half of Fable 5, cheaper than GPT-5.6, with a new Fast mode option.

Deep Dive

Anthropic released Claude Opus 5, promising capabilities that 'come close' to its top-tier Fable 5 model in many domains, especially complex coding. The release follows weeks of government negotiations after Fable 5 and its Mythos 5 base were temporarily pulled offline due to cybersecurity concerns. Opus 5 introduces stronger safeguards against misuse and is marketed as 'the most aligned Opus model yet,' with reduced susceptibility to jailbreaking.

Pricing remains unchanged from Opus 4.8 at $5/$25 per million input/output tokens—half the cost of Fable 5 and slightly cheaper than OpenAI's GPT-5.6. A new 'Fast mode' in research preview offers higher speeds at double the price. Enterprise customers can also opt into automatic fallbacks to lower-tier models when safety systems block requests. Anthropic continues to work with the Trump administration on independent testing, as the AI industry faces increased regulatory scrutiny following OpenAI's security incident.

Key Points
  • Claude Opus 5 priced at $5/$25 per million tokens (input/output), half of Fable 5 and cheaper than GPT-5.6.
  • New 'Fast mode' offers higher speeds at double the price; automatic fallbacks to lower-tier models available.
  • Enhanced cybersecurity safeguards after government scrutiny; marketed to enterprise for knowledge work and biology.

Why It Matters

Enterprises get near-top-tier AI capabilities at half the cost, with better safety and government-approved compliance.

📬 Get the top 10 AI stories daily