AI API for Türkiye: One API for 708+ Models

An AI API is a web service that lets you add large language models such as GPT-5.6 and Claude to your own application with a few lines of code. The mechanics are simple: your application sends an HTTP request containing a model name, messages and optional parameters. The model generates a response and the API returns it as JSON. The unit of measurement is the token — the text you send (input) and the text the model produces (output) are counted as tokens, and billing follows that count.

OpenAI compatibility means, in practice, that the request and response schema matches the OpenAI Chat Completions format that has become the industry standard. A project already using the OpenAI SDK can connect to entirely different models by changing just two parameters (api_key and base_url). That is exactly what Onysoft AI Gateway does:

• One endpoint: https://api.onysoft.com/v1
• One API key for every provider: OpenAI, Anthropic, Google, DeepSeek and more
• Streaming, function calling and vision work the way you are used to
• When a new model ships, only the model parameter changes in your code

What can you build on top of this? The most common scenarios: customer-support bots and in-product chat assistants, question answering over your own documents (RAG), bulk summarization and classification pipelines, coding assistants, and multi-step agent workflows. The basic shape is the same in all of them — send a request, process the response — so learning one compatible interface covers every scenario.

Most global AI providers are headquartered abroad: payment is in USD by foreign card, the paperwork is not a Turkish e-invoice, and support does not speak Turkish. For teams building from Türkiye that adds up to real financial and operational friction. Onysoft AI Gateway removes it:

• Turkish Lira balance: You top up in lira; currency conversion uses the current Turkish Central Bank (TCMB) rate, transparently — no surprise spreads.
• No foreign card required: No need for virtual cards, overseas accounts or VPN workarounds.
• Corporate e-invoices: All spending is documented with Turkish-standard e-invoices, so finance runs its familiar domestic expense process instead of handling foreign service statements.
• KVKK framework: Your contracting party is a Türkiye-based company, and data processing is handled within the framework of KVKK, Türkiye's data protection law. See our KVKK-compliant AI guide for details.
• 24/7 support in Turkish: Technical questions get answers in the team's own language, at any hour.

These advantages matter most to three profiles: startups that must budget in Turkish Lira, agencies and software houses that need to invoice clients for expenses, and enterprise teams that have to clear procurement and legal reviews. For a deeper look at reaching the OpenAI API from Türkiye specifically, see our ChatGPT API in Türkiye guide.

Picking the right model is not about the "strongest" one — it is about the right one for the job. Four questions guide the decision: How much complex reasoning does the task require? How long a context window do you need (a single page or hundreds)? Is response speed or response quality the priority? What per-job cost is sustainable at your volume? Your answers point to one of three tiers:

• Flagship (highest capability): Claude Fable 5 and GPT-5.6 — for complex reasoning, agent workloads and long-document analysis.
• Balanced (price/performance): Claude Sonnet 5 and Gemini 3.6 Flash — the ideal starting point for most production workloads.
• Economical (high volume): GLM, DeepSeek and Kimi K2.x — for classification, summarization and cost-sensitive batch jobs.

You are not limited to text either: image, video and music models run through the same AI API, with the same key. This is where the gateway approach really pays off: switching tiers is a one-parameter change, so there is no such thing as a "wrong" choice — start in the balanced tier, measure on real traffic, then move up or down. Browse the full list in the model catalog, and check the model rankings for independent comparisons.

If you already have an OpenAI integration, switching takes minutes; starting from scratch is just as quick:

1. Create your free account — no foreign card required.
2. Top up your balance in Turkish Lira.
3. Generate your sk-ony- prefixed API key from the dashboard.
4. Change the base_url in your code and send your first request.

from openai import OpenAI

client = OpenAI(
    api_key="sk-ony-YOUR_KEY",
    base_url="https://api.onysoft.com/v1"
)

response = client.chat.completions.create(
    model="claude-sonnet-5",
    messages=[{"role": "user", "content": "Introduce yourself in one sentence."}]
)
print(response.choices[0].message.content)

The request body and response schema are the OpenAI format you already know: streaming, function calling and vision work as-is, and beyond Python you can use the same endpoint from Node.js, Go or plain HTTP. To switch models, change only the model parameter. Endpoints and authentication are covered in the API documentation; to experiment without writing code, the Playground is ready.

AI API usage is billed per token: the text you send (input) and the text the model produces (output) are counted separately, and each model has its own unit price per million tokens. For a rough feel, one token is about 4 characters of English text, and a page of text runs 500-1,000 tokens. On Onysoft the logic stays as simple as possible:

• Pay as you go: No subscription, no commitment; only actual usage is deducted from your balance.
• Turkish Lira balance: Conversion uses the current TCMB rate, so you can budget your spend in lira.
• Per-model pricing: Test with economical models for pennies, then scale up in production as needed.

Three practical ways to keep costs under control: route simple tasks to the economical tier (not every request needs a flagship), keep system prompts short (they are counted on every request), and carry long chat histories forward as summaries. Current unit prices are published in the model catalog; estimate the monthly cost of your own scenario with the cost calculator, and read our AI API pricing article for a detailed breakdown of how pricing works.