Claude API Pricing
Anthropic lists 9 Claude models on its API. Prices below are per million tokens, in US dollars, for input (your prompt), output (the model's reply), cached input (prompt-cache reads) and the batch API.
Every price verified on or after from Anthropic's pricing page (claude.com) and Anthropic's pricing page (platform.claude.com).
Claude API prices per million tokens
| Model | Input | Output | Cached input | Batch input | Batch output | Context |
|---|---|---|---|---|---|---|
| Claude Haiku 5.5 | $0.10$0.50 over 100K | $0.50$2.50 over 100K | $0.01$0.05 over 100K | $0.05$0.25 over 100K | $0.25$1.25 over 100K | 1M |
| Claude Fable 5.1 | $10.00 | $50.00 | $0.25 | $5.00 | $25.00 | 1M |
| Claude Opus 5.5 | $4.00 | $20.00 | $0.20 | $2.00 | $10.00 | 1M |
| Claude Sonnet 5.5 | $2.00 | $10.00 | $0.10 | $1.00 | $5.00 | 1M |
| Claude Opus 4.7 | $5.00 | $25.00 | $0.50 | $2.50 | $12.50 | 200K |
| Claude Haiku 4.5 | $1.00 | $5.00 | $0.10 | $0.50 | $2.50 | 200K |
| Claude Sonnet 4.6 | $3.00 | $15.00 | $0.30 | $1.50 | $7.50 | 1M |
| Claude Sonnet 4.5 | $3.00 | $15.00 | $0.30 | $1.50 | $7.50 | 1M |
| Claude Opus 4.5 | $5.00 | $25.00 | $0.50 | $2.50 | $12.50 | 200K |
A dash means Anthropic lists no such rate for that model. Batch rates apply the batch API discount to the standard rates. A second line shows the higher rate billed for the whole request when its prompt is longer than the stated size, cache reads and writes included. These are Anthropic's direct API rates; cloud platforms can charge differently.
What the headline prices leave out
- Claude Haiku 5.5: Priced by prompt length, per request. Standard (prompt of 100K tokens or less): $0.10 in / $0.01 cached / $0.50 out. Over 100K tokens (counting cache reads and writes), the whole request is billed at $0.50 in / $0.05 cached / $2.50 out. Batch is 50% off both tiers: $0.05 in / $0.25 out standard, $0.25 in / $1.25 out over 100K. Cache writes: 5-minute TTL $0.125/MTok standard and $0.625 over 100K; 1-hour TTL $0.20 standard and $1 over 100K. US-only inference (inference_geo: "us") is billed at 1.1x, including the over-100K prices.
- Claude Fable 5.1: Cache write shown is the 5-minute TTL rate; 1-hour cache writes are $20/MTok. Cache reads are 0.025x input. US-only inference (inference_geo: "us") is billed at 1.1x.
- Claude Opus 5.5: Cache write shown is the 5-minute TTL rate; 1-hour cache writes are $8/MTok. Cache reads are 0.05x input. Fast mode (Claude API only, no batch) is $8 in / $40 out. US-only inference (inference_geo: "us") is billed at 1.1x.
- Claude Sonnet 5.5: Cache write shown is the 5-minute TTL rate; 1-hour cache writes are $4/MTok. Cache reads are 0.05x input. US-only inference (inference_geo: "us") is billed at 1.1x.
Which Claude model is cheapest for your workload
Monthly cost of 10 million tokens on each model, for five common usage patterns. The cheapest model for each pattern is marked. The cheapest model is not always good enough for the task, so compare quality on your own prompts before you switch. Cached workloads assume each cache entry is read 10 times; a model without a batch API pays its standard rate on the batch job.
| Model | Chat assistant7.5M input, 2.5M output | Document Q&A9M input, 1M output | Content generation2.5M input, 7.5M output | Repeated long prompt7.5M input (75% cache hits), 2.5M output | Overnight batch job7.5M input, 2.5M output via the batch API |
|---|---|---|---|---|---|
| Claude Haiku 5.5 | $2.00Cheapest | $1.40Cheapest | $4.00Cheapest | $1.51Cheapest | $1.00Cheapest |
| Claude Fable 5.1 | $200.00 | $140.00 | $400.00 | $146.56 | $100.00 |
| Claude Opus 5.5 | $80.00 | $56.00 | $160.00 | $59.19 | $40.00 |
| Claude Sonnet 5.5 | $40.00 | $28.00 | $80.00 | $29.59 | $20.00 |
| Claude Opus 4.7 | $100.00 | $70.00 | $200.00 | $75.39 | $50.00 |
| Claude Haiku 4.5 | $20.00 | $14.00 | $40.00 | $15.08 | $10.00 |
| Claude Sonnet 4.6 | $60.00 | $42.00 | $120.00 | $45.23 | $30.00 |
| Claude Sonnet 4.5 | $60.00 | $42.00 | $120.00 | $45.23 | $30.00 |
| Claude Opus 4.5 | $100.00 | $70.00 | $200.00 | $75.39 | $50.00 |
Each request is priced on its own prompt size. The patterns above use prompts of at most 9K tokens, so Claude Haiku 5.5 is priced at its standard rates. If your prompts are longer, Claude Haiku 5.5 bills prompts over 100,000 tokens at $0.50 input and $2.50 output per million for the whole request, so these figures understate it.
- Chat assistant: Claude Haiku 5.5 at $2.00 a month.
- Document Q&A: Claude Haiku 5.5 at $1.40 a month.
- Content generation: Claude Haiku 5.5 at $4.00 a month.
- Repeated long prompt: Claude Haiku 5.5 at $1.51 a month.
- Overnight batch job: Claude Haiku 5.5 at $1.00 a month.
Claude Haiku 5.5 is cheapest for every pattern because its input and output rates are both the lowest Anthropic lists for text. Cheapest is not best: the registry has no capability tier, so check quality on your own prompts before you switch.
Price your own workload with Claude Opus 5.5Cite this data
Vanta Calc, Claude API Pricing, https://www.vantacalc.com/technology/llm-pricing/anthropic/, last updated 12 October 2026
Last updated
Reuse licensing is awaiting owner confirmation. For licensing enquiries, use our contact form.