Updated
Updated · CNBC · Jul 24
Anthropic Launches Claude Opus 5 at Half Price, Beating Fable 5 on Coding Benchmarks
Updated
Updated · CNBC · Jul 24

Anthropic Launches Claude Opus 5 at Half Price, Beating Fable 5 on Coding Benchmarks

3 articles · Updated · CNBC · Jul 24

Summary

  • Claude Opus 5 debuted Friday as Anthropic’s top-performing general-use model, priced at $5 per million input tokens and $25 per million output tokens—half the cost of Fable 5.
  • Anthropic said the model beats Fable 5 on coding and knowledge-work evaluations, aiming at enterprises that want everyday AI tools with clearer value for money.
  • That pricing push comes as Anthropic and OpenAI court more cost-conscious customers while facing cheaper offerings from Microsoft, Amazon, Google and Chinese startups.
  • Opus 5 also arrives under continued safety scrutiny after Anthropic’s Mythos and Fable 5 releases were briefly blocked by U.S. export controls; the company said it kept working with government agencies before deployment.
  • Anthropic said Opus 5 still trails Mythos 5 in offensive cybersecurity and biology research, underscoring its positioning as a safer mainstream model rather than the company’s most powerful high-risk system.

Insights

Why did Anthropic drastically reduce Opus 5's safety triggers, and does this signal a retreat from their strict AI alignment principles?
Can Opus 5 truly distinguish between defensive code scanning and malicious exploit generation without accidentally enabling bad actors?
How will Anthropic’s shift from hard safety blocks to controlled degradation reshape the way enterprise applications handle restricted AI prompts?