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