muverqqw/Noir

What the fuck is this model about?

Noir-Standard (3B) is a 3.1 billion parameter model from the Noir series, developed by IceL1ghtning and built on the Qwen 2.5 architecture. It's engineered for high efficiency, aiming to provide advanced logical and mathematical reasoning, typically found in larger models, while being deployable on systems with 8GB of VRAM.

What makes THIS different from all the other models?

This model distinguishes itself by offering a powerful combination of logical reasoning and creative writing capabilities within a compact 3.1B parameter footprint. It achieves a notable 65.0% on GSM8K for mathematical tasks and a 79.7% creativity score for narrative generation. Its ability to maintain stable logic and follow complex multi-step instructions (ARC score of 26.0) makes it a "Smart Professional" assistant that doesn't compromise on creative depth, all while being optimized for local deployment.

Should I use this for my use case?

You should consider Noir-Standard if your use case involves:

• Technical and Logical Problem Solving: Its strong performance on GSM8K makes it ideal for tasks requiring mathematical and logical reasoning.
• Creative Writing and Narrative Generation: The model excels at crafting vivid stories and maintaining distinct character voices over long contexts.
• Resource-Constrained Environments: Designed to fit within 8GB of VRAM, it's perfect for local deployment where efficiency is crucial.
• Multi-step Instruction Following: Its stable logic ensures it can handle complex prompts without losing coherence.

It's a versatile option for developers needing a capable yet lean language model.