Claude Opus 5 API pricing
Claude Opus 5 from Anthropic costs $5.00 per 1M input tokens and $25.00 per 1M output tokens. Cached input drops to $0.50 per 1M. Batch processing takes 50% off both rates.
That puts it 55th cheapest of the 60 models tracked here on a blended 3:1 input-to-output rate, or $10.00 per 1M blended tokens. Verified 2026-08-16.
- Input
- $5.00per 1M tokens
- Output
- $25.00per 1M tokens
- Cached input
- $0.50 90% off input
- Batch discount
- 50% off input + output
- Price rank
- #55of 60, cheapest first
- Tokens per $1
- 100,000at a 3:1 mix
What Claude Opus 5 costs on a real workload
Per-million-token rates are hard to reason about, so here is the same model priced against four concrete monthly workloads. Each row uses this model's own input and output rates against a fixed token mix, with no caching or batch discount applied.
| Workload | Assumption | Monthly cost | Annual |
|---|---|---|---|
| Support chatbot | 50,000 conversations/month at 3K input and 500 output tokens each | $1,375 | $16,500 |
| RAG document search | 200,000 queries/month at 6K retrieved-context input and 400 output tokens | $8,000 | $96,000 |
| Coding agent | 5,000 runs/month at 60K input and 8K output tokens per run | $2,500 | $30,000 |
| Bulk classification | 5,000,000 items/month at 400 input and 20 output tokens each | $12,500 | $150,000 |
Your mix will differ. The API cost calculator takes your own token counts, and the cost per request calculator scales a single call up to per-1,000 and per-month totals.
Cheaper alternatives to Claude Opus 5
These are the closest cheaper options from other providers, ordered by how near they sit to Claude Opus 5 on the blended rate. Closer is usually a more realistic swap: the further down this list you go, the more capability you are likely trading away.
| Model | Provider | Input / 1M | Output / 1M | Cheaper by |
|---|---|---|---|---|
| Sonar Pro | Perplexity | $3.00 | $15.00 | 1.7x |
| Kimi K3 | Moonshot AI | $3.00 | $15.00 | 1.7x |
| GPT-5.4 | OpenAI | $2.50 | $15.00 | 1.8x |
| Nova Premier 1.0 | Amazon | $2.50 | $12.50 | 2.0x |
| Gemini 3 Pro | $2.00 | $12.00 | 2.2x |
Before switching, price the move properly: the model switching savings calculator puts two models against the same workload and shows the annual difference.
Models priced near Claude Opus 5
If cost is roughly fixed and you are choosing on capability instead, these are the models sitting closest to Claude Opus 5 on price.
| Model | Provider | Input / 1M | Output / 1M | Blended |
|---|---|---|---|---|
| Claude Opus 4.8 | Anthropic | $5.00 | $25.00 | $10.00 |
| GPT-5.6 Sol | OpenAI | $5.00 | $30.00 | $11.25 |
| GPT-5.5 | OpenAI | $5.00 | $30.00 | $11.25 |
| Claude Sonnet 4.6 | Anthropic | $3.00 | $15.00 | $6.00 |
How to pay less for Claude Opus 5
Prompt caching. Cache-hit input tokens bill at $0.50 instead of $5.00, which is 90% off. This matters most for workloads that resend a large, mostly-static prefix on every call: system prompts, tool definitions, retrieved documents, long conversation histories. The saving scales with your cache hit rate, so the prompt caching savings calculator is the honest way to size it rather than assuming a perfect hit rate.
Batch processing. Anthropic takes 50% off both input and output when the work goes through the async batch API instead of the real-time one. Anything latency-tolerant qualifies: bulk classification, offline summarisation, dataset labelling, nightly enrichment jobs. The batch API savings calculator shows what that is worth against your volume.
About Anthropic
Anthropic splits Claude into Fable, Opus, Sonnet, and Haiku tiers, and layers a 10% cached-input rate and a flat 50% batch discount on every one of them.
See every Anthropic model and how the lineup is tiered on the Anthropic pricing page, or put Claude Opus 5 against all 60 models from all 17 providers on the comparison table.
Claude Opus 5 pricing FAQ
How much does Claude Opus 5 cost per 1M tokens?
Claude Opus 5 costs $5.00 per 1M input tokens and $25.00 per 1M output tokens. Cached input tokens are cheaper again at $0.50 per 1M, a 90% discount on repeated context.
Is Claude Opus 5 expensive compared to other models?
It ranks 55 of 60 on a blended rate that weights input and output 3:1, so 54 tracked models are cheaper and 5 are more expensive. That works out to 182x the blended rate of Qwen3.7 Flash, the cheapest model tracked here.
What does a real workload cost on Claude Opus 5?
A support chatbot handling 50,000 conversations a month at 3K input and 500 output tokens each comes to about $1,375 a month. A coding agent doing 5,000 runs at 60K input and 8K output per run comes to about $2,500. Run your own numbers in the API cost calculator.
Why is Claude Opus 5 output more expensive than input?
Output on this model is 5.0x the input rate. Every output token needs its own forward pass through the model, while input tokens are processed in parallel, so output costs more to serve across essentially every provider. It also means a workload's input:output mix, not just its total token count, drives the bill.
What is a cheaper alternative to Claude Opus 5?
Sonar Pro from Perplexity is the closest cheaper option at $3.00 / $15.00, roughly 1.7x cheaper on a blended basis. Whether it is a real substitute depends on whether your task actually needs the extra capability.