Smith-3/simon-fcyt-umss

TEXT GENERATIONConcurrent Unit Cost:1Model Size:0.35BQuant:BF16Context Size:32kPublished:Oct 31, 2025License:otherArchitecture:Transformer0.0K Featherless Exclusive Cold

The simon-fcyt-umss model by Smith-3 is a 0.35 billion parameter, fine-tuned version of LiquidAI/LFM2-350M, specifically designed for the TecnoTime application at the Universidad Mayor de San Simón (UMSS). It specializes in generating structured JSON responses for academic reminders and motivational messages, ensuring consistent and parseable output for student engagement applications. Optimized for resource-constrained devices, it provides structured notifications to reinforce academic habits and well-being.

Loading preview...

Overview

Smith-3/simon-fcyt-umss is a 0.35 billion parameter model, fine-tuned from LiquidAI/LFM2-350M, specifically for the TecnoTime application at the Universidad Mayor de San Simón (UMSS). Its core purpose is to generate consistent, validated, and parseable JSON responses for academic support, following the SimonResponse class for Android systems. The model has been converted to GGUF format using Unsloth, making it suitable for deployment on llama.cpp, text-generation-webui, and devices with limited resources like laptops, university labs, and local inference on phones.

Key Capabilities

  • Structured JSON Output: Unlike typical LLMs, this model is constrained to produce a single JSON object with predefined fields, ensuring predictable and machine-readable responses.
  • Academic Support: Generates content for class reminders, motivational check-ins, and positive reinforcement messages tailored for students.
  • Resource Efficiency: Optimized for low-resource environments, offering different GGUF quantizations (Q5_K_M, Q8_0, Q4_K_M) to balance quality and performance.
  • Specific Notification Types: Supports various notification templates like MOTIVATIONAL_CHECK_IN, CLASS_REMINDER, CHECK_IN_CLOSURE, and REMINDER_CLOSURE, each with required and optional fields.

Good For

  • Student Engagement Applications: Ideal for systems like TecnoTime that require automated, structured communication to help students stay organized and motivated.
  • Local Inference on Edge Devices: Its GGUF format and small parameter count make it excellent for deployment on devices with limited computational power.
  • Applications Requiring Structured Output: Any use case where a language model needs to reliably output JSON data rather than free-form text.
  • Reinforcing Academic Habits: Designed to encourage attendance, consistency, and academic well-being through timely reminders and positive messages.

Finetunes
2 models