AI API Providers in Türkiye Compared (2026): Lira Payments, e-Invoices, Model Catalogs

As of 2026, a team that wants to use AI APIs from Türkiye has two main routes: local gateways that accept Turkish Lira and issue local invoices, and global aggregators that stand out for sheer model variety. In this guide we put Onysoft, YapayZekaLab, llm.gen.tr, CollectAPI, OpenRouter, and CometAPI side by side on five concrete criteria: lira payments, corporate e-invoices, model count, OpenAI compatibility, and Turkish-language support. The goal is not to crown a winner but to clarify which provider makes sense for which use case. The information was compiled from each provider's public pages in July 2026; details change over time, so verify the current state on each provider's own site before deciding.

The AI API Market in Türkiye: Two Categories, Five Criteria

Subscribing directly to OpenAI, Anthropic, or Google from Türkiye is technically possible; in practice, foreign-card requirements, USD billing, and foreign service statements that do not fit Türkiye's e-invoice system create real friction. Over the past two years, that friction has grown two categories of solutions. The first is local AI gateways: services that accept Turkish Lira and provide single-point access to multiple model providers. The second is global aggregators: platforms like OpenRouter and CometAPI that unify hundreds of models behind one API but offer no Türkiye-specific payment or invoicing process. We covered the direct-access friction in detail in our ChatGPT API from Türkiye guide; for the bigger picture, see the AI API for Türkiye guide.

We compare on five criteria, because these are the questions that actually decide the purchase:

• Lira payments: For teams budgeting in Turkish Lira, this removes exchange-rate exposure and the need for a foreign card.
• Corporate e-invoices: Decisive for expense booking and tax processes in companies; a foreign service statement means extra work for finance.
• Model count: Shows how broad a catalog one key unlocks — and protects against vendor lock-in.
• OpenAI compatibility: Determines whether existing code migrates with a base_url change.
• Turkish-language support: Determines whether the team gets timely help in its own language when something breaks.

Comparison Table (July 2026)

The table below reflects what each provider's public pages state as of July 2026. Where a cell says "not stated," it does not mean the feature is absent — it means we could not find explicit information about it on the provider's site.

Transparency note: this table was compiled by Onysoft from publicly available sources in July 2026; details may change. Verify the current state on each provider's own site.

We deliberately left out a price column: model prices vary by provider and by month, and a stale price table misleads more than it helps. Current lira pricing for every model in the Onysoft catalog is on the /models page, and the overall pricing logic is covered in our AI API pricing guide.

Provider by Provider: What Each One Offers

Onysoft AI Gateway

An OpenAI-compatible API layer operated by Onysoft Veri Merkezi A.S., headquartered in Izmir, Türkiye. You point your code at https://api.onysoft.com/v1 as the base_url, use an sk-ony- prefixed key, and one balance covers 708+ models including Claude, GPT, Gemini, Grok, DeepSeek, and Kimi. The lira balance is topped up by card or bank transfer/EFT, conversion uses the current Turkish Central Bank (TCMB) rate, spending is documented with corporate e-invoices, and data processing operates within the framework of KVKK, Türkiye's data protection law.

YapayZekaLab

As stated on its site as of July 2026, YapayZekaLab offers access to Claude, GPT, and Gemini models through one account and one balance over an OpenAI-compatible surface, with package purchases or pay-as-you-go and a Turkish-language WhatsApp channel. No fixed model count or corporate e-invoice information is published on the site.

llm.gen.tr

Positioning itself as "Türkiye's AI API service provider," llm.gen.tr lists 30 models on its site as of July 2026 and highlights an OpenAI-compatible API, lira-denominated credit packages, and credits that do not expire. Support runs over email; we could not find explicit e-invoice information on the site.

CollectAPI

CollectAPI really belongs to a different category: a Türkiye-based API marketplace operating since 2018. It offers a broad catalog of data APIs — exchange rates, weather, fuel prices, on-duty pharmacies — and while it has an AI category, it is not an LLM gateway and works with its own endpoint schemas rather than OpenAI SDK compatibility. Worth evaluating for teams that need Türkiye-specific data services rather than LLM access.

OpenRouter

The best known global aggregator. As of July 2026 it serves 400+ models from 70+ providers behind one OpenAI-compatible API. Payment is by USD credit top-ups (card or crypto), with an additional processing fee on credit purchases; current rates are on OpenRouter's own site. There is no Turkish-language support and no Turkish-standard e-invoicing.

CometAPI

A global aggregator that states on its site that a single OpenAI-compatible key unlocks 500+ models, on a pay-as-you-go model, positioning itself on pricing below official list rates. Payment is in USD, and no Türkiye-specific payment method, e-invoicing, or Turkish-language support is offered.

Choosing Guide: Which Provider for Which Scenario?

The right provider depends on your team's constraints. An honest map across the six scenarios we see most often:

• Hobby project or first experiment: To try models without writing code, local providers' trial balances or the Onysoft Playground are practical starting points; OpenRouter's free-model list also works for experimentation.
• Individual developer / freelancer: If you have no foreign card and budget in lira, a local gateway is the natural choice. A small catalog is fine if you only need one model; if variety matters, check catalog breadth first.
• Startup scaling in production: Compatibility, catalog breadth, and cost tracking become critical. If a USD budget works for you, OpenRouter is a strong option; if you need a lira budget, local invoices, and Turkish support, Onysoft covers the same catalog-breadth requirement with local processes. The model rankings and cost calculator help with the comparison.
• Enterprise team / KVKK sensitivity: A Turkish contracting party, corporate e-invoices, and KVKK-framed data processing are preconditions in most procurement processes. Among the providers compared, Onysoft is the one that explicitly commits to all three; see the KVKK-compliant AI guide for details.
• Data API needs: If what you need is not an LLM but Türkiye-specific data services — exchange rates, weather, fuel prices — CollectAPI's marketplace model is built for exactly that.
• Foreign invoicing is not a problem: For teams with USD budgets that can process foreign statements and do not need Turkish support, OpenRouter and CometAPI are legitimate options with wide catalogs.

Where Onysoft Sits in This Table

As the party compiling this comparison, let us summarize our own position with the same honesty. Onysoft's difference is not one feature but the combination: the catalog breadth of a global aggregator (708+ models — Claude Fable 5, GPT-5.6, Gemini 3.6 Flash, Grok 4.5, Kimi K3, DeepSeek V4, and more) joined with a local provider's payment and invoicing processes:

• Lira balance: Topped up by card or bank transfer/EFT; conversion is transparent at the current TCMB rate.
• Corporate e-invoices: All spending is documented to Turkish standards.
• KVKK: Your contracting party is a Turkish company based in Izmir.
• OpenAI compatibility: Works with the official OpenAI SDK; only api_key and base_url change in your code.
• 24/7 support in Turkish: Technical answers in the team's own language, at any hour.

To get started, create a free account, generate your sk-ony- key, and set base_url to https://api.onysoft.com/v1; the details are in the API documentation. Current lira pricing per model is in the model catalog, and the cost calculator compares costs for your scenario. Whichever provider you choose, our advice is the same: base the decision on a small pilot with your own workload, not on any single landing page.