neuracoder/neura-fa-en-1.9b

TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Jun 6, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Neura-FA-EN-1.9B by Neuracoder is an ultra-lightweight 1.9 billion parameter bilingual conversational model built on the Qwen2 architecture, designed for natural, fast, and local conversation in Persian (Farsi) and English. With a 32,768 token context length, it runs efficiently on CPU-only systems and edge devices like Raspberry Pi. This model excels at daily chit-chat, simple Q&A, informal translation, and light summarization, providing an offline assistant for Persian-speaking users.

Loading preview...

Neura-FA-EN-1.9B: A Bilingual Conversational Model for Persian and English

Neura-FA-EN-1.9B is an open-source, ultra-lightweight 1.9 billion parameter conversational model developed by Neuracoder, an Iranian AI company. Built on the Qwen2 architecture, it is specifically designed for natural, fast, and local conversation in both Persian (Farsi) and English. Unlike larger multilingual models, Neura-FA-EN-1.9B is optimized to run smoothly on laptops, CPU-only systems, and even Raspberry Pi, requiring only 4 GB RAM for FP16 inference.

Key Features & Performance

  • Truly Bilingual: Understands and generates both Persian and English fluently, supporting natural code-switching.
  • Ultra-lightweight & Efficient: At 1.9B parameters, it's approximately 1.6 GB (FP16) or 0.9 GB (INT8), enabling fast inference (e.g., 8-12 tok/s on Intel i7 CPU, 2-3 tok/s on Raspberry Pi 4).
  • Long Context: Features a 32,768 token context length, suitable for extended conversations or short stories.
  • Offline & Private: Operates without an internet connection or API key after download, ensuring privacy.
  • Research-Friendly: Released under an Apache 2.0 license to foster community fine-tuning and development.

Evaluations show competitive performance on Persian benchmarks like ParsiMMLU (48.7% accuracy) and PersianQA (56.2% Exact Match), while also performing well on English MMLU (51.3% accuracy). It stands out as the only 1.9B model offering strong performance in both languages.

Suitable Use Cases

  • Daily Chit-chat: Casual conversations, small talk, and friendly assistant tasks.
  • Simple Q&A: Answering general knowledge questions in both languages.
  • Informal Translation: Translating short sentences or phrases between Persian and English.
  • Light Summarization: Summarizing paragraphs or short articles.
  • Brainstorming & Writing Help: Generating ideas or fixing simple grammar.
  • Educational Tool: Practicing Persian or English for language learning.
  • Offline Assistant: Deployment on edge devices for local chatbots or web UIs.

Not suitable for: Code generation, complex mathematical reasoning, professional translation, or tasks requiring up-to-date information beyond mid-2024.

Finetunes
1 models