Claude Opus 5 costs $5 per million input tokens and $25 per million output tokens — the same as Opus 4.8 before it, and half what Claude Fable 5 charges. Those two numbers are the whole headline. Everything else is about how much you actually end up sending.
The full price list
| Mode | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Standard | $5.00 | $25.00 |
| Fast mode | $10.00 | $50.00 |
| Cached input (read) | ~$0.50 | — |
| Batch API | $2.50 | $12.50 |
Cache reads cost roughly a tenth of the base input price; cache writes carry a premium of 1.25x for the five-minute lifetime or 2x for the one-hour option. The Batch API halves both rates for work that can wait.
What that costs in real requests
Token pricing is abstract until you convert it. Rough guide, at standard rates:
| Task | Input | Output | Approximate cost |
|---|---|---|---|
| Question about a 50-page report | ~35K | ~1K | ~$0.20 |
| Review a 500-line pull request | ~15K | ~3K | ~$0.15 |
| Analyse a 750-page document set | ~1M | ~5K | ~$5.13 |
| Long agentic coding session | ~2M | ~200K | ~$15.00 |
The pattern worth internalising: output is five times the price of input, so verbosity is a cost problem, not just a readability problem. An instruction that shortens responses by a fifth cuts a real line item.
Five ways to pay less without changing model
- Cache your prefix. A stable system prompt or document corpus caches from 512 tokens up, and cached reads cost about a tenth of the standard input rate.
- Batch anything that can wait. The Batch API is 50% cheaper and most jobs finish within the hour.
- Tune the effort level. Opus 5 performs unusually well at
lowandmedium; the default ishigh, which is not always the right setting. - Instruct for brevity. Output tokens are the expensive half, and a concrete word limit is far more effective than asking for concision.
- Route, don't upgrade. Send routine work to Sonnet 5 or Haiku 4.5 and reserve Opus 5 for the cases where being right matters.
How it compares
| Model | Input | Output |
|---|---|---|
| Claude Fable 5 | $10 | $50 |
| Claude Opus 5 | $5 | $25 |
| Claude Opus 4.8 | $5 | $25 |
| Claude Sonnet 5 | $3 | $15 |
| Claude Haiku 4.5 | $1 | $5 |
Note that Priority Tier does not cover Opus 5, and it draws on its own rate-limit pool rather than the shared Opus 4.x bucket — worth checking before you move production volume onto it.
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
How much does Claude Opus 5 cost?
$5 per million input tokens and $25 per million output tokens on the Claude API. Fast mode doubles both to $10/$50. The Batch API halves both to $2.50/$12.50, and cached input reads cost roughly a tenth of the standard input rate.
Is Claude Opus 5 cheaper than Claude Fable 5?
Yes, exactly half on both sides: $5/$25 against Fable 5's $10/$50. For most workloads Opus 5 is the better default, with Fable 5 reserved for the hardest long-horizon runs.
What is the cheapest way to run Claude Opus 5?
Combine prompt caching with batch processing. Cached reads cost about a tenth of standard input pricing and the Batch API halves both rates, so a cached, batched workload can land far below the sticker price. Beyond that, route routine requests to a cheaper model entirely.
