BENCHMARKQ2 2026
ZML-Commerce v1.0 achieves 94.2% intent accuracy on Gulf Arabic commerce benchmark
New benchmark results across product discovery, checkout, and order-management intents on a held-out Gulf Arabic commerce test set.
Read updateنماذج الذكاء الاصطناعي لتجارة الخليج
Zameel Labs builds the AI primitives that power the next generation of WhatsApp-native commerce in the GCC.
1
Model Active
ZML-Commerce v1.0
GCC
Native Training
Gulf conversational data
AR / EN
Bilingual Native
with code-switching
01 · Models | النماذج
Conversational commerce model for WhatsApp-native checkout, product discovery, and order management.
نموذج التجارة المحادثاتية عبر واتساب
View Model CardArabic dialect NLU fine-tuned on GCC conversational patterns and Gulf commerce vocabulary.
فهم اللغة الطبيعية لمنطقة الخليج
View Model CardMultimodal product understanding and catalog-to-conversation mapping.
فهم المنتجات متعدد الوسائط
View Model CardUltra-low latency intent classifier. Runs before the larger models as a routing layer: buy, browse, inquire, negotiate, complain, track.
تصنيف النوايا بزمن استجابة فائق السرعة
View Model CardUnderstands and handles bargaining, discount requests, and price negotiation, a commerce behavior deeply embedded in GCC culture that no Western model accounts for.
نموذج المفاوضة والتفاوض على الأسعار
View Model CardPost-purchase conversation layer. Order status, tracking updates, returns, complaints, all handled conversationally over WhatsApp.
إدارة ما بعد الشراء عبر المحادثة
View Model CardDetects fake orders, suspicious conversation patterns, and COD abuse. Trained on GCC-specific fraud patterns, a problem invisible to Western models.
كشف الاحتيال في طلبيات التجارة الخليجية
View Model CardSynthesizes merchant conversation data into intelligence: what products customers ask for but cannot find, peak intent windows, common objections.
تحويل بيانات المحادثات إلى ذكاء تجاري
View Model CardGenerates WhatsApp-native product descriptions and reply templates in Gulf Arabic and English, written the way GCC customers actually read.
كتابة المحتوى التجاري بأسلوب خليجي أصيل
View Model Card02 · Why ZML | لماذا زميل
Trained on Gulf commerce patterns, Arabic dialects, and WhatsApp conversation flows, not generic multilingual datasets.
تدريب على أنماط التجارة الخليجية
Native understanding of how GCC customers actually speak, mixing Arabic and English mid-sentence, mid-order, mid-negotiation.
التبديل الطبيعي بين العربية والإنجليزية
Built for the constraints and patterns of WhatsApp, not adapted from a general-purpose chat model.
بنية مصممة لواتساب أصلاً
Featured Model | النموذج المميز
Capabilities
Intent accuracy on GCC test set
Gulf Arabic dialects supported
Average inference latency
From the Lab | من المختبر
New benchmark results across product discovery, checkout, and order-management intents on a held-out Gulf Arabic commerce test set.
Read updateRequest Access | اطلب الوصول
ابدأ البناء مع نماذج زميل