Opus 5 vs Fable 5: Why the Cheaper Model Is the Right Default

Overview

Claude Opus 5 and Claude Fable 5 share a 1 million token context window, a 128,000 token output limit, and the full low-to-max effort ladder — but Opus 5 costs exactly half as much ($5 and $25 per million tokens versus $10 and $50). Opus 5 also lets you disable thinking, dial effort down, and run under zero data retention, none of which Fable 5 allows. Fable 5 still holds the top capability tier for the hardest reasoning, so the choice is about matching the model to the workload rather than picking the higher rank.

In this video

• Claude Opus 5 costs $5 per million input tokens and $25 per million output; Claude Fable 5 costs $10 and $50 — exactly double on both sides.
• Both models share a 1 million token context window and can emit up to 128,000 output tokens.
• Both support the full effort ladder: low, medium, high, xhigh, and max.
• On Opus 5 thinking is on by default and can be disabled at effort high or below; disabling it at xhigh or max returns a 400 error.
• On Fable 5 thinking is always on by design and cannot be disabled at all — an explicit disable returns a 400 error.
• Fable 5 requires 30-day data retention and is not available under zero data retention, so an organization running ZDR gets a 400 on every Fable 5 request.
• Opus 5 has no data-retention requirement, which makes it the only option of the two for many regulated teams.
• Fast mode runs on Opus 5 but not on Fable 5, for the same model at higher throughput.
• Opus 5 supports mid-conversation tool changes, letting you add or drop a tool between turns without invalidating the prompt cache.
• Claude Fable 5 remains Anthropic's most capable widely released model and holds the top capability tier, so Opus 5 is the better default rather than the outright more capable model.

Frequently asked questions

Is Claude Opus 5 better than Claude Fable 5?

Not in raw capability — Fable 5 remains Anthropic's most capable widely released model and holds the top tier for the hardest reasoning and longest autonomous runs. Opus 5 wins on price, control, latency, and deployability, which makes it the better default for most workloads.

How much do Claude Opus 5 and Claude Fable 5 cost?

Claude Opus 5 is $5 per million input tokens and $25 per million output tokens. Claude Fable 5 is $10 and $50 — exactly twice as much on both sides, for the same 1 million token context window and 128,000 token output limit.

Can you turn off thinking on Claude Fable 5?

No. On Fable 5 thinking is always on by design and an explicit disable returns a 400 error. On Opus 5 thinking is on by default but can be disabled at effort high or below, which matters for fast, cost-sensitive routes.

Can I use Claude Fable 5 under zero data retention?

No. Fable 5 requires 30-day data retention and is not available under zero data retention, so an organization configured for ZDR receives a 400 error on every Fable 5 request. Opus 5 has no such requirement.

When should I choose Fable 5 over Opus 5?

Choose Fable 5 for the hardest unsolved problems where capability is worth almost any price — long overnight migrations and deep research runs. Use Opus 5 for everything else, including agentic coding, code review, and enterprise document work, starting at high effort and sweeping down.