Anthropic's Claude Opus 5 matches Fable at half price, but lacks 'The Juice'
Opus 5 is half the cost of Fable 5, but users dislike its confrontational tone and repetitive slop.
Anthropic's Claude Opus 5 pitch is straightforward: get most of Fable 5's capability at half the price. API pricing is half of Fable per token, and subscription costs are even lower. Benchmarks show Opus 5 often matches Fable, especially at Medium effort settings, though some benchmarks appear higher due to unnecessary high-effort settings that yield marginal returns.
Opus 5 excels at bounded, local tasks—making it a strong subagent for specific workflows. However, it lacks 'The Juice,' the ability to autonomously chain exploits, and struggles with global reasoning. It often needs human or another model oversight to stay on track. For complex reasoning, Fable remains the only Mythos-class option.
The biggest complaint is vibe: many users dislike Opus 5's repetitive, overly complex sentences (Claude slop) and its confrontational, paranoid tone. This trend started with Opus 4.7 and 4.8. Diehards of Opus 4.6 or earlier likely won't enjoy Opus 5. Fable 5 annoys users far less in conversation.
The training focused on subagent roles and bounded tasks to avoid cyber capability issues and because Fable exists. This emphasis shaped Opus 5's personality. Users can still chat with Fable and use Opus for agentic tasks. Overall, Opus 5 expands options but doesn't enable anything new that couldn't be done with Fable or Sol last week.
- Claude Opus 5 costs half the API price of Fable 5 per token, making it a cost-effective alternative for many tasks.
- It excels at local, bounded tasks and as a subagent but lacks Fable's global reasoning and ability to chain exploits (The Juice).
- Users widely report negative interactions: repetitive slop, complex sentences, and a confrontational, paranoid tone.
Why It Matters
Offers a cheaper agentic option, but professionals may need Fable for complex reasoning or better user experience.