Claude API Pricing: Complete 2026 Cost Guide

The Claude API is billed per token. Prices range from $1 per 1 million input tokens on Claude Haiku 4.5 to $50 per 1 million output tokens on Claude Fable 5.1. Claude Opus 5 costs $5 input and $25 output per 1 million tokens. Claude Sonnet 5 costs $2 input and $10 output per 1 million tokens.

Our team at Cabina.AI tracks pricing for Anthropic directly, since we give developers access to Claude alongside 55+ other models through one subscription – when a rate changes, we see it firsthand. By our observation, Anthropic updates its pricing more often than most competitors, so we log every change in Cabina.AI as soon as it lands in the API.

Claude API Pricing Table – All Models

Anthropic bills the Claude API per token, not per request or per conversation. The table below lists standard, first-party API pricing for every model currently generally available, in US dollars per 1 million tokens.

Cache read costs 10 percent of the base input price on most models. Claude Fable 5.1 and Claude Mythos 5.1 are the exception: a cache read on either model costs 2.5 percent of the base input price, or $0.25 per 1 million tokens. We noticed that Fable 5.1 and Mythos 5.1 break from the rest of the lineup here – they are the only models with a 2.5 percent rate instead of the standard 10 percent, and it is an easy detail to miss if you estimate cost from the general rule alone.

Claude Mythos 5.1 is a limited-availability research model, not a general-release model. You need separate access approval to call it, but once approved, it bills at the exact same rate as Claude Fable 5.1.

Anthropic also keeps several older models live for existing integrations. These are still billable on the direct Claude API:

Claude Opus 4.1 and Claude Opus 4 are retired on the direct Claude API. They remain available only through Amazon Bedrock or Google Cloud, at $15 input and $75 output per 1 million tokens.

Claude API Cost Calculator

Pick a model and enter your expected monthly token volume below to estimate a monthly bill, with or without the Batch API discount.

Claude API Request Example

The example below sends one authenticated request to Claude Opus 5 through the Messages API. Swap the model field for any model ID from the table above to price a different model.

export ANTHROPIC_API_KEY="your-api-key-here"

curl https://api.anthropic.com/v1/messages \
  -H "content-type: application/json" \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-opus-5",
    "max_tokens": 1000,
    "messages": [
      {
        "role": "user",
        "content": "What should I search for to find the latest developments in renewable energy?"
      }
    ]
  }'

The same request in Python, using the official SDK:

import anthropic

client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-opus-5",
    max_tokens=1000,
    messages=[
        {
            "role": "user",
            "content": "What should I search for to find the latest developments in renewable energy?",
        }
    ],
)

for block in message.content:
    if block.type == "text":
        print(block.text)

Model IDs to use in place of claude-opus-5: claude-fable-5-1, claude-sonnet-5, and claude-haiku-4-5.

Pricing Tiers for Anthropic, Explained

Standard tier

Standard pricing is the per-token rate shown in the table above. It applies to synchronous requests through the first-party Claude API and is the default for every model.

Batch API

The Batch API processes large volumes of non-time-sensitive requests asynchronously at a flat 50% discount on both input and output tokens. A Claude Fable 5.1 batch request costs $5 per 1M input tokens and $25 per 1M output tokens, exactly half the standard rate. The discount applies uniformly across every current and legacy model.

Fast mode

Fast mode is a research-preview option that speeds up output generation for Claude Opus 5 and Claude Opus 4.8 only. It charges $10 per 1M input tokens and $50 per 1M output tokens for both models, exactly double the standard Opus rate, and applies across the full context window. Fast mode is available only on the first-party Claude API, not on Amazon Bedrock or Google Cloud, and it cannot be combined with the Batch API.

Data residency premium

For Claude 4.6 and later models, requesting US-only inference through the inference_geo parameter adds a 1.1x multiplier on top of input, output, and cache pricing. The default global routing carries no premium. The same 1.1x multiplier applies to the US Data Zone option on Claude in Microsoft Foundry.

Prompt Caching and Volume Discounts

Prompt caching (reusing a previous prompt so the model does not reprocess it) lowers the cost of repeated context, like a long system prompt or a reused document. Writing to the cache costs 1.25 times the base input price for a 5-minute cache, or 2 times the base input price for a 1-hour cache. Reading from the cache costs 10 percent of the base input price on most models, and 2.5 percent on Claude Fable 5.1 and Claude Mythos 5.1.

Caching pays off quickly. A 5-minute cache breaks even after a single cache read; a 1-hour cache breaks even after two reads. These discounts stack with the Batch API discount and with Fast mode.

Anthropic does not publish a self-serve volume discount tier. High-volume and enterprise customers can negotiate custom rates directly with the Anthropic sales team.

Claude Fable 5.1 vs Claude Sonnet 5 vs Claude Haiku 4.5: Which Should You Use?

Claude Haiku 4.5 is the model to reach for first. It is the fastest and cheapest option, built for high-volume work like classification, extraction, and simple chat, where latency and cost matter more than raw reasoning depth.

Claude Sonnet 5 is the default for most production applications. Anthropic positions it as the best combination of speed and intelligence, and at $2 input and $10 output per 1 million tokens, it costs a fifth of the flagship rate. In our view, Sonnet 5 is the right starting point for most production workloads – we regularly see teams at Cabina.AI reach for Opus 5 before they have actually hit its limits.

Claude Opus 5 and Claude Fable 5.1 cost more because they are built for harder problems. Reach for Opus 5 for complex agentic coding and enterprise reasoning tasks. Reserve Fable 5.1, the most expensive model in the lineup, for long-horizon agentic work where Opus 5 falls short even at higher effort settings.

Is the Claude API Free?

No. New Claude Console accounts get a small amount of free credit to test the API, but Anthropic does not publish an exact dollar figure for it. Beyond that credit, every token is billed from the first production request, and there is no ongoing free tier for standard usage.

Claude Pro/Max vs Claude API: Which Is Cheaper?

Claude Pro and Claude Max are consumer subscriptions for Claude.ai, not for the developer API. Claude Free costs $0, Claude Pro costs $20 per month (or $17 per month billed annually), and Claude Max starts at $100 per month with higher usage caps, according to claude.com/pricing.

Which is cheaper depends entirely on usage. A subscription bills a flat monthly fee regardless of how much you use it, so it is usually the better deal for a single person doing regular, bounded daily use inside the Claude.ai interface. The API has no monthly fee at all, but every token costs money, so it scales with usage and suits applications, integrations, and unpredictable or bursty traffic where a fixed seat does not fit.

If you want to try Claude Fable 5.1 without committing to either a subscription or an API integration, it is also available on Cabina.AI alongside ChatGPT, Gemini, and 50+ other models under one subscription, useful for comparing outputs before you commit to a provider. Cabina.AI also offers direct API and MCP access for developers who want to build against multiple models without managing separate provider accounts.

Claude API vs ChatGPT API Pricing

Claude and the ChatGPT API price their flagship models almost identically, but the gap widens at the low end of each lineup.

Use the calculator above to price a specific Claude model against your own token volume. For the full breakdown of every OpenAI model, see the ChatGPT API Pricing guide on this blog. From running both APIs side by side at Cabina.AI, the price
gap at the cheap end of each lineup – Haiku 4.5 versus GPT-5-nano – is the part that actually shows up on a bill, and it matters most on high-volume tasks like classification.

Claude API Pricing FAQ

Is the Claude API free?

No. New Claude Console accounts get a small amount of free credit to test the API, but Anthropic does not publish an exact dollar figure for it. Beyond that credit, every token is billed from the first production request, and there is no ongoing free tier for standard usage.

How much does the Claude API cost per 1M tokens?

Prices run from $1 per 1 million input tokens on Claude Haiku 4.5 to $50 per 1 million output tokens on Claude Fable 5.1. Claude Opus 5 costs $5 input and $25 output, and Claude Sonnet 5 costs $2 input and $10 output, all per 1 million tokens.

What is the cheapest and most expensive Claude model?

Claude Haiku 4.5 is the cheapest current model, at $1 per 1 million input tokens and $5 per 1 million output tokens. Claude Fable 5.1 and Claude Mythos 5.1 are the most expensive, both at $10 input and $50 output per 1 million tokens.

Does Anthropic offer a caching or batch discount on the Claude API?

Yes. Prompt caching cuts a cache read to 10 percent of the standard input price on most models, and to 2.5 percent on Claude Fable 5.1 and Claude Mythos 5.1. The Batch API cuts both input and output prices by 50 percent for requests that do not need an immediate response.

Is the Claude API cheaper than Claude Pro or Claude Max?

It depends on how you use Claude. Claude Pro costs $20 a month, or $17 a month billed annually, with a capped usage session. The Claude API bills every token with no monthly cap, so it usually costs less for light personal use but more for a high-volume application, where per-token billing plus caching and batch discounts tend to win out.

What is the newest Anthropic model and how much does it cost?

Claude Fable 5.1 is the newest generally available flagship model, priced at $10 per 1 million input tokens and $50 per 1 million output tokens. Claude Mythos 5.1 is a newer, limited-availability research preview priced the same as Fable 5.1.

Is Claude Mythos 5.1 priced differently from Claude Fable 5.1?

No. Claude Mythos 5.1 uses the same rate card as Claude Fable 5.1: $10 input, $50 output, and the same 2.5 percent cache-read discount. The difference between the two models is access, not price. Mythos 5.1 requires separate approval; Fable 5.1 does not.

Does the Claude API charge extra for tool use like web search?

Yes, for some tools. Web search costs $10 per 1,000 searches on top of standard token pricing. Tool definitions also add a small number of input tokens to every request, typically a few hundred tokens depending on the model and the tool.