AI Inference Built for Egypt
An OpenAI-compatible inference API for Egypt. Billed in EGP. Sovereign path for regulated workloads.
Works with the tools you already use
LlamaIndex
Hermes AgentOne API. Every model.
Chat, embedding and multimodal models from the live catalog — EGP rates, context windows, and modalities, all behind one OpenAI-compatible endpoint.
Your data is never stored
This model runs on GPUs inside Egypt. Your prompts and outputs are processed in memory and never saved, and they are never used to train models. Requests to this model are served in-country.
Qwen3.8-27B-FP8
Alibaba
Input / 1M
14.0EGP
Output / 1M
105EGP
Context
131k ctx
Your data is never stored
This model runs on GPUs inside Egypt. Your prompts and outputs are processed in memory and never saved, and they are never used to train models. Requests to this model are served in-country.
Qwen3-0.6B
Alibaba
Input / 1M
5.00EGP
Output / 1M
10.0EGP
Context
4k ctx
claude-opus-4.8
Anthropic
Input / 1M
294EGP
Output / 1M
1.5kEGP
Context
1M ctx
gpt-5.5
OpenAI
Input / 1M
294EGP
Output / 1M
1.8kEGP
Context
1.1M ctx
gemini-3.1-pro
Input / 1M
118EGP
Output / 1M
706EGP
Context
1M ctx
deepseek-v4-pro
DeepSeek
Input / 1M
76.5EGP
Output / 1M
153EGP
Context
1.0M ctx
MoonshotAI: Kimi K3
moonshotai
Input / 1M
168EGP
Output / 1M
838EGP
Context
1.0M ctx
qwen3-embedding-8b
Alibaba
Input / 1M
0.59EGP
Output / 1M
Free
Context
33k ctx
Local currency. No dollar card. No FX fees.
Pay-as-you-go from prepaid EGP credit, topped up by card or mobile wallet. Most rates are set from our providers’ USD prices at the current exchange rate, so they can change — current rates are always on the pricing page and in the dashboard.
meta-llama-3.1-8b-instruct-turbo
Meta · chat
Input / 1M
1.18EGP
Output / 1M
2.35EGP
Est. monthly*
94EGP
Qwen: Qwen3.5 Plus 2026-04-20
Alibaba · chat
Input / 1M
17.6EGP
Output / 1M
106EGP
Est. monthly*
2,470EGP
embeddinggemma-300m
Google · embed
Input / 1M
0.12EGP
Output / 1M
Free
Est. monthly*
5.9EGP
* Illustrative monthly total for 50M input + 15M output tokens on chat models (embeddings bill input only). Amounts in EGP.
See pricing for every modelThree lines of code.
Any chat model.
Invoiced in EGP.
Get your API key
Sign up and verify your email (a personal organization is created for you), then create an API key from the dashboard.
Pick a model
Browse the catalog with EGP pricing and context windows side-by-side.
Ship in EGP
Call it with the OpenAI SDK for chat and embeddings. Use prepaid credits from your account balance.
from openai import OpenAI
client = OpenAI(
base_url="https://backend.sovereigneg.com/v1",
api_key="sk-...",
)
response = client.chat.completions.create(
model="gpt-oss-20b", # any live chat model id from the catalog
messages=[
{"role": "user", "content": "Hello!"}
]
)
print(response.choices[0].message.content)Works with the tools you already use.
SovereignEG is OpenAI-compatible. Change three values — the base URL, the API key and the model name — and your framework or agent app runs on SovereignEG models. Every guide is short, tested, and step by step.
Enterprise AI controls, built in.
Orgs, roles, spend caps and per-project request logs are in the product today, not on a roadmap slide. Enterprise pilots can request our written security review.
Organizations & projects
Two-tier hierarchy mirroring OpenAI: orgs own billing, projects own keys and usage. In team orgs, extra projects can be limited to a model allow-list.
Members & roles
Invite teammates by email into a team org with preset roles (Owner / Member / Viewer); Owners can also define custom roles. Roles are checked server-side for API-key management, playground use and org, project and member administration.
API keys with guardrails
Per-key monthly EGP spend caps, RPM limits and optional expiry. Keys are shown once and can be revoked instantly.
Request logs
Inference calls are logged per project (model, status, latency, tokens, EGP cost), and your team can review the last 30 days in the dashboard.
EGP-native billing
Pay-as-you-go from a prepaid EGP credit balance. Self-serve top-up by card or mobile wallet through Kashier, coupon redemption, transaction history per org, and a PDF invoice for every successful payment.
Per-project spend caps
Optional monthly EGP cap per project (Cairo calendar month), checked before every inference request, with a clear 402 error once it is reached. Requests already running when the cap is reached still finish and are billed, so spend can end above the cap.
From cloud API to sovereign AI hosting in Egypt — one product.
Same API surface, same EGP billing path. Customers don’t re-integrate when their workload moves sovereign.
Standard routing — live today
Available nowOne OpenAI-compatible endpoint, billed in EGP. Most models are served by vetted upstream providers outside Egypt; models marked “Hosted in Egypt” in the catalog are served first from our own GPU in Egypt. Best for product teams shipping today.
Unified API
One base URL and API key for chat, embeddings and image generation. OpenAI SDK code for Chat Completions and Embeddings works after changing the base URL, the key and, where needed, the model ID; the Responses, Assistants, Audio, Files and Batch APIs are not offered.
Live model catalog
Browse every available model with per-1M-token EGP pricing in the dashboard.
EGP credit balance
Prepaid credits. Top up by card or mobile wallet through Kashier; credit lands automatically once the payment is confirmed.
Sovereign deployment — Egypt-hosted
Design partnersDedicated GPU capacity hosted inside Egypt for regulated workloads. Currently in private design-partner phase.
Egypt-hosted capacity
Dedicated, per-customer deployments inside Egypt for approved design partners, separate from the shared standard route.
Private deployment
Isolated capacity per customer; sized and procured per design-partner contract.
Designed for PDPL
Designed to support Egypt Personal Data Protection Law and sector-specific data-residency rules.
Two ways to start.
Developers self-serve. Regulated enterprises talk to us first.
AI access today. Enterprise governance now. Dedicated sovereign hosting next.