Enterprise AI with Muse & Llama in Turkey - The Unified AI API Platform

What Is Meta Muse & Llama?

About the technology and the company behind the model

Meta — the company behind Facebook, Instagram and WhatsApp — kicked off the open-weight revolution in AI when it released the first Llama model in 2023. The family has since passed a billion downloads, making it the largest open-model ecosystem in the world and proof that frontier-class AI does not have to live behind a single vendor's closed API.

Alongside the open-weight Llama family, Meta is building a new model series called Muse. The Muse Wonder series, announced on October 12, 2026, is in Meta's own words the company's most capable models to date; Meta positions them for personal and professional agentic tasks, with examples such as processing insurance claims, reviewing leases and building apps. Muse Wonder is available to developers through an API and is in the Onysoft catalog as meta/muse-wonder; the larger Muse Wonder Full is, for now, offered only to subscribers of Meta's Muse Code and cannot be reached through an API. The earlier Muse Spark 1.3 is in the catalog too.

The current Llama 4 generation raises the bar with a Mixture-of-Experts (MoE) architecture. The strongest of the series, Llama 4 Maverick, activates just 17 billion of its 400 billion total parameters per request, delivering high quality at a fraction of the compute — and it understands images natively alongside text. Efficiency-focused Llama 4 Scout stretches the context window to as much as 10 million tokens, enough to process entire document archives in one call, while the battle-tested Llama 3.3 70B remains the industry workhorse for high-volume production workloads.

Onysoft AI Gateway, operated by Izmir-based Onysoft Veri Merkezi A.S., makes Llama API access from Turkey effortless: top up a Turkish lira balance and pay as you go at the central bank's current exchange rate with live local pricing, receive corporate e-invoices, and rely on KVKK-compliant processes backed by 24/7 support in Turkish. Your existing OpenAI SDK code works with nothing more than a base_url change, and the same API puts 769 models next to Llama for easy side-by-side testing — no foreign credit card or VPN required.

Open-weight models are fast becoming the default for organizations that want cost control and data sovereignty. As Meta ships new Muse and Llama releases, they land in the Onysoft catalog with up-to-date TL pricing.

business

Provider

Muse & Llama Performance Metrics

Results from industry-standard benchmark tests

MMLU Pro Multi-Domain Knowledge (Llama 4 Maverick) 80.5

GPQA Diamond Scientific Reasoning (Llama 4 Maverick) 69.8

LiveCodeBench Code Generation (Llama 4 Maverick) 43.4

ChartQA Chart & Visual Understanding (Llama 4 Maverick) 90.0

* Benchmark results are based on data published by the respective providers and may be updated.

Why Choose Muse & Llama?

Meta's Muse Models and Open-Weight Llama Family

lock_open

Open-Weight Freedom

Llama's weights are public: no vendor lock-in, full transparency. Start on the API and keep the option to move the exact same model to your own infrastructure later.

psychology

MoE Efficiency

Llama 4's Mixture-of-Experts architecture activates only a small slice of its total parameters per request. The result: high quality at noticeably lower latency and cost.

description

Massive Context Window

Llama 4 Scout stretches to 10 million tokens of context while Maverick handles 1 million — enough to process entire archives or book-length documents in one request.

image

Natively Multimodal

Llama 4 models are trained with early-fusion multimodality to understand text and images together. Send screenshots, charts and document photos straight to the API.

code

Strong Coding & Tool Use

Solid performance on code generation, debugging and refactoring, with function calling and JSON output support that makes it agent- and automation-ready.

payments

Pay As You Go in Lira

No subscriptions on Onysoft: top up in Turkish lira and pay only for what you use at the central bank's current rate. Open-weight economics meet live local pricing.

Muse & Llama Models and Pricing

Muse & Llama models available in TL through Onysoft AI Gateway

Recommended

Muse Wonder

New

The API model of the Muse Wonder series that Meta announced on October 12, 2026 and presents as its most capable models to date; built for personal and professional agentic tasks. 1,048,576-token context in the catalog.

meta/muse-wonder play_circle Try

arrow_downward Input: ₺0,30 / 1K tokens

arrow_upward Output: ₺1,50 / 1K tokens

memory Context: 1,048,576 tokens

Muse Spark 1.3

The earlier Muse model: multimodal reasoning for long-running agent and coding work. 1M-token context.

meta/muse-spark-1.3 play_circle Try

arrow_downward Input: ₺0,09 / 1K tokens

arrow_upward Output: ₺0,32 / 1K tokens

memory Context: 1,048,576 tokens

Llama 4 Maverick

Open Weights

128-expert MoE: 400B total / 17B active parameters, natively multimodal, 1M-token context. The strongest of the open-weight Llama family, for complex reasoning and enterprise workloads.

meta-llama/llama-4-maverick play_circle Try

arrow_downward Input: ₺0,03 / 1K tokens

arrow_upward Output: ₺0,08 / 1K tokens

memory Context: 524,288 tokens

Llama 4 Scout

Efficiency-focused MoE: 109B total / 17B active parameters with up to 10M tokens of context. Ideal for huge documents and cost-sensitive, high-volume jobs.

meta-llama/llama-4-scout play_circle Try

arrow_downward Input: ₺0,01 / 1K tokens

arrow_upward Output: ₺0,02 / 1K tokens

memory Context: 1,310,720 tokens

Llama 3.3 70B

Popular

The proven open model: 128K context and broad ecosystem support. The dependable standard for high-volume production workloads.

meta-llama/llama-3.3-70b-instruct play_circle Try

arrow_downward Input: ₺0,04 / 1K tokens

arrow_upward Output: ₺0,06 / 1K tokens

memory Context: 131,072 tokens

view_list View All Models and Pricing

Muse & Llama Use Cases

What businesses in Turkey can build with Muse & Llama

smart_toy

High-Volume Chatbots

Serve millions of customer-service and in-app assistant messages at low cost — Scout covers the speed tier while Maverick handles the quality tier.

search

Document Analysis & RAG

Analyze contract archives, technical documentation and knowledge bases without chunking thanks to Scout's 10M-token context — and simplify your RAG pipeline.

code

Code Assistant

Code completion, automated code review and CI error analysis — economical enough to run on every commit thanks to open-weight pricing.

analytics

Data Extraction & Classification

From unstructured text to JSON: extract fields from invoices, forms and emails, and label or classify large datasets in batch.

language

Translation & Localization

Localize product content, support articles and marketing copy into dozens of languages, powered by Llama's broad multilingual training data.

model_training

Synthetic Data & Distillation

The open license lets you improve your own smaller models with Llama outputs — ideal for synthetic data generation, distillation and evaluation workflows.

Which Sectors Does Muse & Llama Serve?

Sectors using Muse & Llama through Onysoft AI Gateway

Technology Startups E-Commerce Customer Service Fintech Gaming Media & Publishing Telecom

How to Get Started with Muse & Llama?

Start using Muse & Llama models in your project in 3 steps

1

Create an Account

Sign up to Onysoft AI Gateway for free. Sign in instantly with your email address or your GitHub/Google account. Your API key is generated automatically.

2

Add Balance

Add as much balance as you like in Turkish Lira. You can pay by bank transfer/EFT, credit card or corporate invoice. Transparent TL pricing with the current TCMB exchange rate.

3

Start Using the API

Thanks to our OpenAI-compatible API endpoint, access Muse & Llama models without changing your existing code. Just change the base_url and api_key.

from openai import OpenAI

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

response = client.chat.completions.create(
    model="meta/muse-wonder",
    messages=[{"role": "user", "content": "Merhaba!"}]
)
print(response.choices[0].message.content)

import OpenAI from 'openai';

const client = new OpenAI({
    apiKey: 'sk-ony-...',
    baseURL: 'https://api.onysoft.com/v1'
});

const response = await client.chat.completions.create({
    model: 'meta/muse-wonder',
    messages: [{role: 'user', content: 'Merhaba!'}]
});
console.log(response.choices[0].message.content);

curl https://api.onysoft.com/v1/chat/completions \
  -H "Authorization: Bearer sk-ony-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "meta/muse-wonder",
    "messages": [{"role": "user", "content": "Merhaba!"}]
  }'

$ch = curl_init('https://api.onysoft.com/v1/chat/completions');
curl_setopt_array($ch, [
    CURLOPT_HTTPHEADER => [
        'Authorization: Bearer sk-ony-...',
        'Content-Type: application/json'
    ],
    CURLOPT_POSTFIELDS => json_encode([
        'model' => 'meta/muse-wonder',
        'messages' => [['role' => 'user', 'content' => 'Merhaba!']]
    ]),
    CURLOPT_RETURNTRANSFER => true,
    CURLOPT_POST => true
]);
$response = json_decode(curl_exec($ch));
echo $response->choices[0]->message->content;

Why Use Muse & Llama with Onysoft?

The advantages of choosing Onysoft AI Gateway over the Meta API directly

currency_lira

Pay in Turkish Lira

No foreign credit card needed. Transparent pricing in TL with the current TCMB exchange rate. Bank transfer, credit card or corporate invoice.

receipt_long

Invoicing in Turkey

As a company based in Turkey, we issue e-invoices and e-archive invoices. Your accounting processes run smoothly.

hub

One API, All Models

Access 769+ models including Muse & Llama just by changing the base_url. OpenAI SDK compatible.

support

24/7 Support in Turkish

Our Turkish-speaking technical support team is always with you. Integration help and issue resolution.

shield

KVKK Compliant

Data processing policies aligned with Turkish regulations. Enterprise data security standards.

trending_down

Cost Optimization

Switch easily between different models as your needs change. Optimize your budget.

Frequently Asked Questions About Muse & Llama

What people wonder about Muse & Llama and Onysoft AI Gateway

How do I access the Llama API from Turkey? expand_more

Sign up for Onysoft AI Gateway, top up a Turkish lira balance and you are running within minutes. No foreign credit card, VPN or overseas billing address is needed — just point your client at https://api.onysoft.com/v1. You can also try the models in the Playground without writing any code.

How does Llama API pricing work in Turkish lira? expand_more

There are no subscriptions or commitments — you pay only for the tokens you use. Charges are converted at the Turkish central bank's current exchange rate and deducted from your lira balance, and each model's live local price is shown right on this page.

Is using Llama through Onysoft compliant with KVKK, Turkey's data protection law? expand_more

Yes — Onysoft AI Gateway is operated by Izmir-based Onysoft Veri Merkezi A.S. under KVKK-compliant processes. Your account and billing data are handled under Turkish regulations, and the 24/7 support team assists with contractual and compliance requirements for enterprise deployments.

How do I choose between Muse Wonder, Muse Spark and the Llama models? expand_more

Try Muse Wonder, which Meta presents as its most capable models, for complex, multi-step agentic tasks. Where you want open-weight economics and a low unit cost, Llama 4 Maverick covers quality and the long-context Llama 4 Scout covers volume (Meta states up to 10 million tokens of context for Scout; the Onysoft catalog lists it at 1,310,720 tokens). If you want a proven, stable model with broad ecosystem support, Llama 3.3 70B is the safe choice. Prices differ widely, so compare them with your own prompts in the Playground.

What is Muse Wonder, and can I use it through an API? expand_more

Muse Wonder is the API-available model of the Muse Wonder series that Meta announced on October 12, 2026; Meta presents the series as its most capable models to date. On Onysoft you call it as meta/muse-wonder with your existing OpenAI SDK code, and the catalog lists it with a 1,048,576-token context. The larger Muse Wonder Full is offered only with Meta's Muse Code subscription and cannot be reached through an API.

Does my existing OpenAI SDK code work with Llama? expand_more

Yes. Set base_url to https://api.onysoft.com/v1, use your Onysoft API key and pass a model id such as meta-llama/llama-4-maverick. Python, Node.js and every other OpenAI-compatible client works without further changes.

What are the context window limits of the Llama models? expand_more

Meta states a context of up to 10 million tokens for Llama 4 Scout; in the Onysoft catalog Llama 4 Scout is listed at 1,310,720 tokens, Llama 4 Maverick at 1,048,576 and Llama 3.3 70B at 131,072. That is enough to process hundreds of pages of contracts or an entire codebase in a single request.

What makes Llama 4 Maverick special? expand_more

Maverick uses a 128-expert MoE design that activates only 17B of its 400B total parameters per request, which is how it delivers high quality at a much lower cost. It is also natively multimodal, understanding images alongside text in the same request.

Can I get corporate invoices for business use? expand_more

Yes — every balance top-up comes with a corporate e-invoice issued to your company in Turkish lira. Since billing is fully local, your accounting team avoids foreign-currency reconciliation and cross-border service paperwork.

Llama is open source — why use an API at all? expand_more

Self-hosting Llama 4 Maverick requires multi-GPU H100 servers and ongoing maintenance. Through the API you scale in seconds with zero infrastructure investment — and because the weights are open, you always keep the option of moving the very same model to your own servers later.

How good is Llama at Turkish? expand_more

The Llama 4 generation was trained on far broader multilingual data than earlier releases and handles most everyday Turkish tasks well. For critical content that demands top-tier Turkish quality we still recommend reviewing outputs — you can quickly benchmark it against other models with your own prompts in the Playground.

hub Meta Muse & Llama