Voltar aos guias
API·8 de junho de 2026·Atualizado em 3 de setembro de 2026·7 min de leitura

Claude API — the complete guide to calling Claude on Kunavo

Claude is Anthropic's frontier family for reasoning, coding and agents. This is the hub for calling it on Kunavo — keys, SDKs, models, pricing and caching, all linked.

Last reviewed on .

Claude is Anthropic's frontier family — the default choice for reasoning, coding and agentic workflows. Kunavo serves the whole family, from Haiku to the new Claude 5 generation (Fable 5, Opus 5, Sonnet 5), through both an OpenAI-compatible API and the native Anthropic Messages API, at 3060% under Anthropic's list price depending on the model, on one Stripe-billed balance. This is the hub for everything Claude on Kunavo.

What you can do with the Claude API

  • Reasoning & analysis — long-context document analysis, multi-step problem solving.
  • Coding — generation, refactoring and review; Sonnet and Opus are strong agentic coders.
  • Vision — image understanding alongside text.
  • Tool use — function calling for agents and structured output.
  • Prompt caching — cache large stable context and pay 10% of the input rate on cache hits.

Get a key and make your first call

Create a key at /app/keys, then keep the OpenAI SDK and point base_url at Kunavo. Step-by-step (and the native Messages API form) is in how to get a Claude API key.

claude.py
import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["KUNAVO_API_KEY"],
    base_url="https://api.kunavo.com/v1",
)

resp = client.chat.completions.create(
    model="claude-sonnet-4-6",
    messages=[{"role": "user", "content": "Explain the CAP theorem to a backend engineer."}],
)
print(resp.choices[0].message.content)

Prefer Anthropic's native shape, with cache_control and extended thinking? The same key works against the Messages API (/v1/messages).

Claude models, and when to use each

Rates are per 1M tokens in USD as billed on Kunavo, with Anthropic's published list price alongside. Worked cost examples and the prompt-caching math are in the Anthropic API pricing guide.

ModelKunavo in / outAnthropic listOff listUse it for
claude-haiku-4-5$0.40 / $2.00$1.00 / $5.00~60%High-volume classification, support, simple extraction
claude-sonnet-4-6$1.20 / $6.00$3.00 / $15.00~60%The balanced value default for reasoning, RAG and coding
claude-sonnet-5$2.00 / $10.00$2.00 / $10.00none — same as listNear-Opus coding and agentic quality at Sonnet cost
claude-opus-4-7$2.00 / $10.00$5.00 / $25.00~60%Heavy reasoning
claude-opus-5$2.00 / $10.00$5.00 / $25.00~60%Hard reasoning on a budget — same rate as Opus 4.7, half of Fable 5's list price
claude-fable-5$7.00 / $35.00$10.00 / $50.00~30%Frontier reasoning, long-horizon agents, 1M context

Pricing

Claude on Kunavo runs 3060% under Anthropic's list: the Claude 4.x generation is 60% off (Sonnet 4.6 at $1.20/$6.00 per 1M vs $3.00/$15.00), Fable 5 is 30% off at $7.00/$35.00, and Opus 5 is 60% off at $2.00/$10.00 vs $5.00/$25.00. Sonnet 5 is the one exception: at $2.00/$10.00 it matches Anthropic's list price exactly rather than undercutting it, because Anthropic made that rate permanent on August 31, 2026. Full per-model rates, worked cost examples and the prompt-caching math are in the Claude API pricing guide.

Go deeper

FAQ

How do I get a Claude API key?

Create a Kunavo account, generate a key at /app/keys, and use it with base_url=https://api.kunavo.com/v1. See the full guide.

How much does the Claude API cost?

On Kunavo, Claude runs 30–60% under Anthropic's list price: Haiku 4.5 $0.40/$2.00, Sonnet 4.6 $1.20/$6.00, Opus 4.7 $2.00/$10.00 and the new Opus 5 at the same $2.00/$10.00 (60% off list), and Fable 5 $7.00/$35.00 per 1M tokens (30% off list). The one exception is Sonnet 5 at $2.00/$10.00, which is Anthropic's own list price rather than a discount on it. Prompt caching bills cached input at 10% of the input rate.

Can I use the Anthropic SDK or Messages API?

Yes — both the OpenAI-compatible endpoint and the native Messages API (/v1/messages) work on the same key.

Which Claude model should I use?

claude-sonnet-4-6 is the balanced value default for most reasoning, RAG and coding, with claude-sonnet-5 the newest near-Opus coder. Use claude-haiku-4-5 for cheap high-volume tasks and claude-opus-5 for hard reasoning on a budget — near-flagship quality at $2.00/$10.00 per 1M, half of Claude Fable 5's list price and the same rate as the older claude-opus-4-7, so the upgrade costs nothing extra. Reserve claude-fable-5-1 — Anthropic's most capable model, and the September 2026 successor to claude-fable-5 at the same price — for frontier reasoning and long-horizon agents.