KiwiMate/KiwiMate-Mini-Preview
KiwiMate/KiwiMate-Mini-Preview is a 3.2 billion parameter Llama-based conversational language model developed by KiwiMate / KyleCodeKiwi. Fine-tuned for the KiwiMate AI companion app, it specializes in New Zealand cultural context, Te Reo Māori vocabulary, and app-specific lore. This model is optimized for fast, low-cost inference, providing general-purpose chat with a distinct "Kiwi" flavour.
Loading preview...
KiwiMate-Mini-Preview: A "Kiwi-Flavoured" Conversational Model
KiwiMate-Mini-Preview, developed by KiwiMate / KyleCodeKiwi, is a 3.2 billion parameter Llama-based conversational language model. It is the smallest in the KiwiMate family, designed for efficient, low-cost inference within the KiwiMate AI companion app. This model is currently a prototype, with future updates and behavioral shifts expected before a stable v1 release.
Key Capabilities & Features
- New Zealand Context: Fine-tuned to understand and incorporate New Zealand cultural nuances, slang, and geography.
- Te Reo Māori: Provides light recognition and usage of Te Reo Māori vocabulary.
- App-Specific Lore: Incorporates "KiwiMate Origin" data and knowledge for in-app mini-games.
- Optimized for Inference: Designed for speed and cost-effectiveness, making it suitable for free and lower-tier subscription plans.
- Apache 2.0 License: Released under an open-source license, consistent with the KiwiMate model family's commitment.
Intended Use Cases
- General-purpose chat and assistant-style conversations within the KiwiMate app.
- Providing New Zealand cultural and contextual information.
- Supporting in-app mini-games with relevant knowledge.
Limitations
As a 3B-parameter model, its reasoning depth and factual reliability are limited compared to larger models, prioritizing speed and personality. It is not intended for high-stakes advice (medical, legal, financial) or as an authoritative source on Māori language or tikanga.