Why is Claude Opus 4.5 so expensive?

Claude Opus 4.5 prices near $15 per million input tokens and $75 per million output tokens. Output is the killer: agents that loop through many tool calls generate enormous output volume. A single 30-step agent run can cost more than a full day of Sonnet chat.

The fix is not to avoid Opus. It is to stop paying full price for the parts of the prompt that never change.

How does prompt caching work on Anthropic’s API?

You mark stable content — usually a system prompt and reference docs — with a cache breakpoint. The first call writes the cache; subsequent calls read it at a large discount (roughly a tenth of the input price) as long as the content is unchanged and the cache has not expired.

1. Move all fixed instructions, style guides and reference files to the top of the message array.
2. Add a cache breakpoint after the stable block using the cache control field.
3. Keep variable content — the user’s actual question — after the breakpoint.
4. Send requests within the cache TTL (a few minutes) to get the discount.

If you include a timestamp or random ID in the cached block, you bust the cache every call and pay full price. That mistake alone wiped out the savings in my first attempt.

How much can caching actually save?

On a document-Q&A workload with a 20,000-token system prompt reused across 500 calls, caching dropped input cost by about 88% on the cached block. The output tokens stayed the same, so the total bill fell by roughly a third.

For agent loops with stable tool definitions, savings were higher because the schema block was large and untouched.

How to combine caching with model routing

Add a cheap classifier in front. If the request is simple — formatting, lookup, short rewriting — send it to Haiku. If it needs real reasoning across many files, send it to Opus. This two-layer setup is what most teams should run.

What else trims the bill?

Trim context aggressively. Do not paste the whole repo when three files will do — every extra token is paid on every call unless cached. Cap agent iteration counts so a stuck loop cannot run forever. And log token usage per request so a regression is visible the day it appears, not at month-end.

When is Opus 4.5 still worth it?

When the task genuinely needs the strongest reasoning and tool use — refactors, complex debugging, long-horizon agents. For everything else, routing to a cheaper tier costs you almost nothing in quality and a lot less in money.

Quick Summary

• Claude Opus 4.5 costs about $15/M input and $75/M output tokens.
• Prompt caching discounts the stable portion of a prompt by up to ~90%.
• Never put timestamps or random IDs inside the cached block — it busts the cache.
• Route simple tasks to Haiku or Sonnet and reserve Opus for hard reasoning.
• Log token usage per request to catch cost regressions early.

I ran it on a 60,000-line TypeScript monorepo for a week. It shipped working diffs more often than any model I have used, but it is not cheap and it is not always the right pick.

What is Claude Opus 4.5?

Claude Opus 4.5 is Anthropic’s flagship model released in late 2025. It supports a 200K-token context window (available on the API, with larger beta options), extended thinking mode, and native tool use including parallel function calls.

API pricing sits around $15 per million input tokens and $75 per million output tokens. That is the highest tier in the lineup — Sonnet and Haiku cost a fraction of it.

How do you access Claude Opus 4.5?

• Claude apps — Pro ($20/month) and Max ($100-200/month) plans include Opus access with usage caps.
• API — model string claude-opus-4-5 via the Anthropic Console or any compatible gateway.
• IDE integrations — Claude Code, Cursor and similar tools let you select Opus as the backing model.

Expect tighter rate limits than Sonnet. Heavy sets of parallel agent tasks will hit the ceiling quickly on lower tiers.

How to structure prompts for hard coding work

1. Give the repo context first. Paste the relevant files or use a tool that provides them. Opus reasons better with real code than with descriptions of code.
2. State the invariant. “Do not change the public API of this module” prevents it from refactoring things you did not ask about.
3. Ask for a plan before the diff. On multi-file changes, request a short plan, approve it, then have it implement. This cut my rework rate roughly in half.
4. Enable extended thinking for architecture questions, disable it for quick edits to save latency.

Where does Claude Opus 4.5 actually excel?

Long-horizon agentic tasks. It can run a sequence of tool calls — read file, run tests, fix failure, re-run — without losing the thread. On a legacy refactor across 40 files it finished with tests passing, something cheaper models abandoned halfway.

It also writes unusually readable code. Comments are sparse but accurate, and it avoids the over-abstracted patterns some models love.

What are the downsides?

Cost is the obvious one. A single large repo-wide refactor can burn several dollars in tokens. Latency with extended thinking on is slow — 30-90 seconds for a plan on a big task.

And it still gets long arithmetic wrong occasionally, so keep a linter and test suite in the loop rather than trusting the raw diff.

Who should use it?

Teams doing serious refactors, agents and multi-file changes get clear value. Solo developers on small scripts should use Sonnet or Haiku and save both money and time.

Quick Summary

• Claude Opus 4.5 is Anthropic’s flagship coding model with a 200K-token context window.
• Access via Claude Pro/Max apps, the API model string claude-opus-4-5, or IDE tools.
• Ask for a plan before a multi-file diff to cut rework.
• Excellent at long agentic refactors; expensive at roughly $15/M input and $75/M output tokens.
• Keep tests and linters in the loop — it still makes occasional arithmetic mistakes.