Claude API Pricing 2026 — Model Rates, Pix Payments, and Actual Costs

Claude is the default for reasoning, code, and agents—and Kunavo offers every Claude model 30% below Anthropic list prices, depending on the model, through a single OpenAI-compatible API. As of September 24, 2026, this includes claude-sonnet-5: Anthropic made the $2.00 / $10.00 rate the permanent standard price on August 31, 2026, and Kunavo is 30% lower. This guide brings together what Brazilian developers search for most about Claude API pricing and payment: current rates by model, how to pay from Brazil (Pix included), cost examples you can verify yourself, and the biggest ways to lower your bill.

Claude API prices at a glance

Rates in USD per 1M tokens, exactly as charged by Kunavo. The “Anthropic list price” column shows the official price for the same model for a direct comparison.

Live rates are always available on the pricing page and on each model page (claude-sonnet-4-6, claude-haiku-4-5, claude-opus-4-7). Haiku is the low-cost workhorse, Sonnet is the balanced default, and Opus and Fable are for heavier reasoning. Claude Opus 5.5, released by Anthropic on September 22, 2026, is the newest Opus: a 1M-token context window and up to 128K output tokens.

Claude Opus 5.5 is the best value for money. Anthropic lists it at $4.00 / $20.00 per 1M tokens—below Opus 5 at $5.00 / $25.00—and says it performs at the level of Claude Fable 5.1 for most work. On Kunavo, it costs $2.80 / $14.00, ~30% below Anthropic list price and less than Opus 5 ($3.50 / $17.50): it is Kunavo’s least expensive Opus. Sonnet 5 costs even less: $1.40 / $7.00, ~30% below list price. A 50K input / 2K output analysis that costs $0,42 on Fable 5 costs $0,21 on Opus 5 and $0,168 on Opus 5.5—and since Opus 5.5 adaptive thinking is always on, reserve some of the output budget for it.

How to pay for the Claude API in Brazil—Pix, cards, Apple Pay

Payment is processed by Stripe. There is in subscription or Anthropic Console setup: you top up a balance, and each call is charged at the per-token rates above. For customers paying from Brazil:

• Pix—instant, with the amount shown in reais at checkout;
• International cards: Visa, Mastercard, American Express;
• Apple Pay, Google Pay, and Link (one-click payment);

Per-token rates are in USD; with Pix, the conversion to BRL appears immediately, while card conversion follows the issuer’s rate (check with your bank about IOF and foreign transaction fees). Minimum top-up: $10, prepaid, and the balance never expires—no monthly fee and in end-of-month surprises. If your card is declined, check whether international purchases are enabled in your banking app—or pay by Pix, which does not depend on that. The Pix process, screen by screen, is covered in the guide to paying for the API with Pix.

Cost examples

Real figures calculated using Kunavo rates:

Processing 10,000 support tickets per day on Haiku costs about $12,60/day. Here is the calculation so you can run it yourself:

# Tarifas Claude na Kunavo (USD por 1M de tokens): (input, output)
RATES = {
    "claude-haiku-4-5":  (0.70, 3.50),
    "claude-sonnet-4-6": (2.10, 10.50),
    "claude-opus-4-7":   (3.50, 17.50),
    "claude-opus-5":     (3.50, 17.50),
    "claude-opus-5-5":   (2.80, 14.00),
}

def custo(model: str, in_tokens: int, out_tokens: int) -> float:
    i, o = RATES[model]
    return in_tokens / 1_000_000 * i + out_tokens / 1_000_000 * o

print(custo("claude-haiku-4-5", 800, 200))      # um ticket de suporte  -> $0.00126
print(custo("claude-sonnet-4-6", 6_000, 500))   # uma resposta RAG      -> $0.0179
print(custo("claude-opus-4-7", 50_000, 2_000))  # análise longa         -> $0.21
print(custo("claude-opus-5", 50_000, 2_000))    # análise no Opus 5     -> $0.21
print(custo("claude-opus-5-5", 50_000, 2_000))  # análise no Opus 5.5   -> $0.168

Portuguese uses ~10–20% more tokens than English for the same content—factor that into your budget.

Prompt caching—the biggest single saving

If your system prompt or retrieved context is large and stable, prompt caching charges cached input at 10% of the input rate (2,5% no Claude Fable 5.1 e 5% no Claude Opus 5.5). For long, repeated prompts, this cuts input costs by 60–90%—just add one field to the request. See the in-depth prompt caching guide and the Messages API reference.

One key for Claude, GPT, images, and video

A single Kunavo key (sk-kn-…) calls Claude, GPT, and image, video, and audio models—all in OpenAI-compatible format and charged to the same Stripe balance, with in need to reconcile invoices by provider. Claude is served through both the compatible endpoint /v1/chat/completions and the native Anthropic Messages API (/v1/messages), so you can keep using your existing SDK. The step-by-step instructions for creating a key are in the authentication documentation.

Which Claude model should you choose?

Code patterns for routing by difficulty and cutting costs are covered in the cost optimization guide. For a comparison with OpenRouter, see Kunavo vs. OpenRouter.

Claude in VS Code—Kilo Code, Cline, Roo Code

If you use a coding agent such as Kilo Code, Cline, or Roo Code, point the “OpenAI Compatible” provider to https://api.kunavo.com/v1 with your Kunavo key and choose a Claude model—three fields and nothing else. A typical agent session uses thousands of tokens per step, so savings of 30%, depending on the model, add up quickly. The step-by-step instructions with screenshots are in the Kilo Code + Claude API guide.

How to reduce Claude API costs

1. Start with prompt caching. Cached input costs 10% of the rate (2,5% no Claude Fable 5.1 e 5% no Claude Opus 5.5)—the biggest single return on investment.
2. Move down a tier. Send easy work to Haiku and reserve Opus/Fable for the difficult fraction.
3. Limit output. Set max_tokens—output costs ~5x as much as input.
4. Trim the input. Retrieve fewer, more precise RAG chunks instead of the entire knowledge base.

Frequently asked questions