arXiv:2605.28827v1 Announce Type: cross 
Abstract: Open Arabic large language models split into two classes: sub-1B multilingual models that treat Arabic as an afterthought (Qwen2.5-0.5B, Falcon-H1-0.5B), and 7B-70B Arabic-specialized models that require a server to run (Jais, AceGPT, ALLaM, SILMA). The one published attempt at a sub-2B Arabic-specialized model, Kuwain-1.5B, never released its weights. We present RightNow-Arabic-0.5B-Turbo, a 518M-parameter Arabic-specialized decoder LLM built on Qwen2.5-0.5B. The pipeline adds 27,032 Arabic tokens via mean-subtoken initialization, continues pretraining on 504M Arabic tokens on 8xH100 with FSDP, FlashAttention varlen packing, and Liger fused kernels, then applies supervised fine-tuning on 129,116 Arabic instruction pairs with response-only loss masking, direct preference optimization on 6,750 Arabic preference pairs, and weight soup merging across three checkpoints. On three lm-evaluation-harness Arabic benchmarks (COPA-ar, Arabic HellaSwag, ArabicMMLU) the merged model reaches 35.9% mean accuracy, beats every same-class open model, ties Falcon-H1-1.5B on COPA-ar (58.4%) at one-third the size, and recovers 67% of SILMA-9B's mean at 1/18 the parameters. The edge build quantizes to 398 MB (q4_k_m) and delivers 635 tokens/s at batch size 1 on a single H100 via llama.cpp. All code (5,555 lines across 25 scripts), weights (bf16, int8, and four GGUF quantizations), and benchmark scripts are released at https://huggingface.co/RightNowAI/RightNow-Arabic-0.5B-Turbo.

تم تقديم RightNow-Arabic-0.5B-Turbo كنموذج لغة عربية متخصص يحتوي على 518 مليون معلمة، مبني على بنية Qwen2.5-0.5B. يهدف هذا النموذج إلى معالجة الفجوة في معالجة اللغة العربية من خلال توفير خيار أكثر كفاءة وسهولة مقارنة بالنماذج الأكبر التي تتطلب موارد خادم كبيرة.

RightNow-Arabic-0.5B-Turbo se ha presentado como un modelo de lenguaje árabe especializado con 518 millones de parámetros, construido sobre la arquitectura Qwen2.5-0.5B. Este modelo busca abordar la brecha en el procesamiento del lenguaje árabe al proporcionar una opción más eficiente y accesible en comparación con modelos más grandes que requieren recursos significativos del servidor.

RightNow-Arabic-0.5B-Turbo a été introduit comme un modèle de langue arabe spécialisé avec 518 millions de paramètres, construit sur l'architecture Qwen2.5-0.5B. Ce modèle vise à combler le fossé dans le traitement de la langue arabe en offrant une option plus efficace et accessible par rapport aux modèles plus grands qui nécessitent des ressources serveur importantes.

RightNow-Arabic-0.5B-Turbo has been introduced as a specialized Arabic language model with 518 million parameters, built on the Qwen2.5-0.5B architecture. This model aims to address the gap in Arabic language processing by providing a more efficient and accessible option compared to larger models that require significant server resources.