Models
Models built for service, native in Arabic.
A generic chatbot is trained to sound plausible. Our models are built to resolve — tuned on real customer-service work, native in Arabic including Gulf dialect, and fluent in English. Purpose beats generality when the goal is an accurate outcome.
Purpose over generality
Not a generic chatbot with an Arabic setting.
Most models treat Arabic as a translation target — English reasoning, then a conversion at the edge. Ours reason in Arabic natively, understand Gulf dialect and Modern Standard Arabic, and switch to English the moment the customer does.
~66%
of conversations resolved with no human
Models tuned for resolution and accuracy — measured on real customer-service conversations, not benchmark trivia. The rest hand off with full context.
What we tune for
The qualities that make a service model, not a chat model
Arabic-native
Understands and reasons in Arabic — including Gulf dialect and Modern Standard Arabic — and code-switches with English mid-conversation.
Tuned for resolution
Optimised for ending conversations with a real outcome, not for the longest or most confident-sounding reply.
Accuracy first
Rewarded for being right and grounded, penalised for plausible-but-wrong — the opposite of a model tuned only for fluency.
Safe by construction
Trained to stay on-policy, refuse what it should, and defer to a human when the stakes or the uncertainty are high.
Service-shaped
Tuned on the shape of real support, sales and ecommerce conversations — refunds, orders, claims, escalations — not open web chat.
On-brand tone
Adapts register and politeness to your brand and to the customer’s language, formality and mood.
Why it matters
The model is the difference between a demo and a deployment.
- Native Arabic understanding means fewer misreads, fewer escalations, and answers that sound like a person, not a translation.
- Tuning for resolution and accuracy keeps confident-but-wrong answers out of production.
- Because grounding and citations are expected of the model, trust is verifiable on every turn.
- One model family serves service, sales and ecommerce, so behaviour stays consistent across every role.
Questions
Common questions
Do you support Gulf dialect specifically?
Yes. The models understand Gulf dialect and Modern Standard Arabic natively, and handle Arabic–English code-switching within a single conversation.
How is this different from a general-purpose chatbot?
General chatbots optimise for sounding plausible across any topic. Our models are tuned specifically for customer-service resolution and accuracy, grounded in your knowledge and measured on real conversations.
Can the model be shaped to our brand?
Yes. Tone, formality and persona adapt to your brand and to each customer’s language and mood, while staying inside the guardrails you set.
Hear the difference a purpose-built model makes.
First month free, up to 10,000 resolved conversations. Live in under a day.