Claude Opus 4.8 API Pricing: Cost Calculator & Real Examples (2026)
Claude Opus 4.8 charges $5.00 per 1 million input tokens and $25.00 per 1 million output tokens. But per-token prices don't answer the real question: what will your workload cost per month? Below we compute it for three realistic scenarios — or run your own numbers with Claude Opus 4.8 preselected.
| Meter | Price per 1M tokens |
|---|---|
| Input | $5.00 |
| Cached input | $0.50 |
| Output | $25.00 |
Note: Cache read = 0.1× input; cache writes billed extra (1.25× for 5-min).
What does Claude Opus 4.8 cost in practice?
| Monthly scenario | Token bill |
|---|---|
| Support chatbot (150k requests, 400 in / 300 out tokens) | $1,425.00 |
| Document workflow (22k runs, 3,000 in / 400 out tokens) | $550.00 |
| High-volume classifier (1M requests, 300 in / 10 out tokens) | $1,750.00 |
In the chatbot scenario, output is ~79% of the Claude Opus 4.8 bill — capping response length (max tokens) is the most direct optimization. Document workflows are input-heavy, so the input rate dominates there.
Prompt caching savings
Claude Opus 4.8 offers cached input at $0.50/1M (vs $5.00 standard). At a 60% cache-hit rate, the example chatbot drops from $1,425.00 to $1,263.00/month — saving $162.00. If your system prompt and context are stable, this discount is free money.
Is Claude Opus 4.8 cheap or expensive?
On the same chatbot workload, Claude Opus 4.8 is cheaper than 2 of the 13 models we track (position 12 of 14, cheapest to priciest). The nearest cheaper alternative is GPT-5.4 ($825.00/month on the same scenario). One step up: GPT-5.6 Sol ($1,650.00/month). The right decision isn't absolute price but the cheapest model that meets your quality bar — test before overpaying.
FAQ
How much does Claude Opus 4.8 cost per 1M tokens?
$5.00 per 1M input tokens and $25.00 per 1M output tokens (verified 2026-07-09).
What does a chatbot cost on Claude Opus 4.8?
A 500-active-user chatbot (10 questions/day, 400 input + 300 output tokens per request) costs ~$1,425.00/month in tokens on Claude Opus 4.8. Adjust the assumptions in the calculator.
Our take on this model
Opus 4.8 is Anthropic's top model and the one enterprises pick when the output is the product: complex code, long legal or technical documents, agent systems that run unattended. Its long-context reliability is the differentiator you're actually paying for.
Cost-wise, Opus punishes chatty architectures. Its economics work when you send fewer, bigger, well-prepared requests — batch your context, cache your prefix, and don't use Opus to answer what Haiku can.
Prices verified on 2026-07-09 against the provider's official pricing page. Estimates are for planning purposes only — always confirm current pricing with the provider.