The AI assistant
built for the Gulf.
Private, fast, and fluent in Arabic, Hindi, and English. It searches the live web, remembers you, reads photos and voice notes, and gets real work done. Free to start, $5/month for the Agent plan.
An assistant that actually does the work.
Chat on the web or Telegram, in Arabic, Hindi, or English. It searches the live web, remembers what matters to you, understands photos and voice, and takes real actions, all on our cheap Gulf power. Free to start.
Searches the live web
Ask anything current and it searches the web, then answers with sources you can click. No stale guesses, no made-up facts.
Remembers you
On the Agent plan it keeps your preferences, projects, and the people you deal with, across every chat. You never have to repeat yourself.
Sees, hears, reads
Send a photo, record a voice note, or drop a file. It reads images, transcribes speech, and understands your documents.
Gets things done
It sets reminders, follows up, and emails notes to you, and always shows you the action and waits for a tap before it runs.
Speaks your language
Fluent in Arabic, Hindi, and English, switching mid-sentence the way the Gulf actually talks.
Everywhere you are
Use it in the browser, put it on Telegram in a minute, or install it as an app on your phone.
- Chat, live web search, photos and voice
- Reminders, follow-ups, and email notes
- Arabic, Hindi, and English
- Everything in Free
- Persistent memory across every chat
- Heavier, hands-off autonomous work
Sovereignty, for the teams that need it.
Regulators across the Gulf now point to one answer for safe AI: run it in-region, on infrastructure you control. That is exactly how the DesertGrid enterprise tier is built, and in the assistant, any chat can be pinned to in-region serving with one tap.
- In-regionTurn on Keep my data in-region in the assistant and that chat is served from a Gulf-hosted model on Liwa desert power, so it stays where your laws apply.
- Never trainedWe never train on your data. Privacy-first by default, aligned with UAE PDPL and GDPR expectations.
- No-logsThe enterprise sovereign tier runs on dedicated instances with no request logging and customer-managed encryption keys.
- Self-hostFor regulated and enterprise teams, run DesertGrid on infrastructure you control, so client data never leaves the building.
Also an OpenAI-compatible API and a model marketplace.
Beyond the assistant, DesertGrid is a developer platform. Call efficient flash-class models over an OpenAI-compatible API on cheap Gulf power, or list your own model on the marketplace. Change two lines and ship.
Cheaper, more private, and yours to brand.
The same flash-class intelligence the big providers sell, on a structurally lower cost base, with terms built for builders rather than billboards.
Built on cheap desert power
Inference cost is dominated by electricity. We run on UAE desert energy at $0.10/kWh through our Liwa infrastructure, a cost floor most providers cannot touch.
Privacy-first, PDPL-native
We never train on your data. The sovereign tier runs no-logs on dedicated instances, and the roadmap keeps inference inside the Gulf for true data residency.
No hard rate limits
Fair-use throttling protects quality, not your wallet. No surprise caps mid-launch, and dedicated capacity the moment you need guaranteed throughput.
Drop-in OpenAI-compatible
Change the base URL and the key. Your existing OpenAI SDK code, agents and tools keep working, no rewrite, no lock-in.
GCC-bound low latency
Serving the region from the region. As we migrate onto UAE infrastructure, Gulf users get the short ping that distant clouds cannot offer.
White-label ready
Resell under your own brand. Your customers, your prices, your portal, our cheap-power engine quietly underneath.
If you can call OpenAI, you can call DesertGrid.
Get a key
Sign in and we issue an API key scoped to your account.
Point your client at our base URL
Keep your OpenAI SDK. Set the base URL to DesertGrid and pick a model.
Ship
Same request and response shapes. Pay per token, no hard limits, watch your cost drop.
from openai import OpenAI client = OpenAI( base_url="https://api.desertgrid.ai/v1", api_key="dg-your-key", ) resp = client.chat.completions.create( model="deepseek-v4-flash", messages=[{"role": "user", "content": "Marhaba from the Gulf"}], ) # same shapes you already use. that's it.
Efficient flash-class models, priced to move.
We serve mixture-of-experts models that fire only a fraction of their parameters per token, the reason quality keeps climbing while price keeps falling. Roster grows over time.
$0.28 / 1M output
$0.14 / 1M output
for agents and RAG
Model availability and pricing track the open-model market and are confirmed at beta onboarding. OpenAI-compatible endpoints for chat, completions and embeddings.
Pay per token, or prepay and save.
For builders calling the API directly. (The assistant is free, with a $5/month Agent plan, above.) Transparent per-model pricing that undercuts the major low-cost providers, because our power bill is lower, not because we are burning runway.
- No commitment, no minimums
- No hard rate limits
- All flash-class models
- OpenAI-compatible API
- Better effective rate as you scale
- Usage dashboard and alerts
- Priority queue at peak
- White-label add-on available
- Dedicated instances, guaranteed throughput
- No-logs, PDPL data residency
- Customer-managed encryption
- Full white-label and SLA
Per-token rates vary by model and by input versus output, as shown in the lineup above. Credit packs are prepaid balances spent at those rates. Prices may change as the open-model market moves.
The OpenRouter of the Gulf.
Soon, anyone will be able to list and monetize their own hosted model on DesertGrid. Bring a model or a GPU, plug into the cheapest power on the grid, and earn on every token served. We take a small platform fee and handle billing, keys, and routing.
Create your account and start building.
Passwordless sign-in with a magic link. Your account is created the first time you log in. Early users get founder pricing, free tokens for feedback, and a direct line to the team.
- ✓ Founder pricing locked for early cohorts
- ✓ Free tokens in exchange for feedback
- ✓ White-label and sovereign options on request
Sign in to buy credits, manage API keys, configure hosting and reserve GPUs.
Sign in to the console →New here? Signing in creates your account, no password needed.
List a model insteadQuestions, answered.
Can the assistant search the web?
Yes. The assistant searches the live web and answers with clickable sources, so you get current information instead of stale guesses. Web search is free.
Does the assistant remember me?
On the $5/month Agent plan it keeps durable memory of your preferences, projects and the people you deal with, across every chat, so you never repeat yourself. You can see and delete anything it has stored.
How much does the Agent plan cost?
The assistant is free to use. The Agent plan is $5 per month and unlocks persistent memory across every chat, the premium model, and heavier, hands-off autonomous work. It is a simple monthly subscription, cancel anytime, no contracts.
Can I run DesertGrid on my own infrastructure?
Yes. For regulated and enterprise teams the sovereign tier runs on dedicated, no-logs instances, or on infrastructure you control, keeping data residency in the Gulf so your data never leaves your control. Talk to us about a sovereign deployment.
Is DesertGrid OpenAI-compatible?
Yes. Point any OpenAI client at our base URL and swap your API key. No SDK rewrite, change two lines and ship.
Where does my data go?
By default the assistant runs on our fastest model so you get the best answers, and we never train on your data. If you need data residency, turn on "Keep my data in-region" to pin that chat to a Gulf-hosted model, and the enterprise sovereign tier runs in-region on dedicated, no-logs instances or on infrastructure you control.
How is it so cheap?
Two reasons. We serve efficient mixture-of-experts flash models that activate only a fraction of their parameters per token, and we are built on ultra-low-cost UAE desert power at $0.10 per kWh through our Liwa infrastructure. Cheap power is a structural cost advantage, not a promotion that expires.
Can I bring my own model?
That is the marketplace. Hosts will be able to list and monetize their own models on our cheap-power infrastructure and earn on every token served, while DesertGrid handles billing, keys and routing for a small platform fee.
Are there hard rate limits?
No hard caps by default. We use fair-use throttling to protect quality, and the sovereign tier gives you dedicated capacity with guaranteed throughput.
How do I get started?
DesertGrid is in early access. Sign in to get started, with founder pricing and free tokens for early users.