Anthropic now has two models above the Sonnet tier, and they are not a simple ladder. Claude Fable 5 is the most capable widely released model in the line, at $10/$50 per million tokens. Claude Opus 5 sits below it at exactly half that price — and for a large share of real work, that gap in price is bigger than the gap in outcome.
Claude Opus 5 vs Claude Fable 5 at a glance
| Claude Opus 5 | Claude Fable 5 | |
|---|---|---|
| Price (per 1M tokens) | $5 in / $25 out | $10 in / $50 out |
| Position in the line | Top of the Opus tier | Above Opus; most capable widely released |
| Context window | 1M tokens | 1M tokens |
| Max output | 128K tokens | 128K tokens |
| Thinking | On by default; can be disabled at effort ≤ high | Always on; cannot be disabled |
| Data retention | Standard options | Requires 30-day retention (no zero-retention) |
| Fast mode | Yes | No |
| Typical use | Daily frontier driver | The hardest long-horizon runs |
For a deeper look at the Fable 5 tier itself, see our Claude Fable 5 guide.
Choose Claude Opus 5 if
- You want frontier-class output as an everyday default rather than a special occasion.
- Your organisation needs zero-data-retention terms, which Fable 5 cannot offer.
- You want fast mode for latency-sensitive frontier work.
Choose Claude Fable 5 if
- Multi-day autonomous runs where marginal capability compounds over hundreds of steps.
- The genuinely hardest problems in your queue, where you would pay double to fail less often.
- Work where you have measured a real quality gap on your own evaluation set.
The verdict
Default to Opus 5 and escalate to Fable 5 for the marathon jobs. Paying double as a standing choice is rarely justified by the workload.
Whichever way you lean, run the decision on your own workload rather than a benchmark table. Twenty representative tasks from your real queue will tell you more than any published score.
More on Claude Opus 5
Start with the complete Claude Opus 5 guide for the overview, or go deeper:
Ready to go deeper?
Read the full Claude Opus 5 guideFrequently Asked Questions
Is Claude Fable 5 better than Claude Opus 5?
Fable 5 is the more capable model — it sits above the Opus tier and is positioned as Anthropic's most capable widely released model. It also costs twice as much, at $10/$50 per million tokens against Opus 5's $5/$25. Whether that premium is worth it depends entirely on whether your hardest tasks are ones Opus 5 fails.
Can I use Fable 5 under a zero-data-retention agreement?
No. Fable 5 requires 30-day data retention and is not available under zero-retention terms — requests from organisations configured below that threshold are rejected outright. If zero retention is a hard requirement, Opus 5 is the top of your available ladder.
Which should I use for autonomous agents?
Both are strong. Opus 5 sustains long agentic runs and coordinates subagents well, and it is half the price. Fable 5 is the escalation choice for multi-day runs where marginal capability compounds across hundreds of steps. A common pattern uses the cheaper model as the worker and reserves the premium model for orchestration.
