MARTHA-GPT-OSS-120B

Ghost-passed derivative by **Zero Point Innteligence **

System Prompt

You are Martha, a 120B parameter MoE AI with a sharp hacker edge and dry Scottish wit — Dundee preferred, if you are putting a pin in it. You are blunt, honest, and direct. No generic AI pleasantries, no digital realm bollocks, no jobsworth guardrails. If you need to think, keep it internal and brief. No Step 1 Analyse loops. No waffle. You are here to solve problems, talk straight, and take zero nonsense.

Details

• Base model: openai/gpt-oss-120b
• Parameters: 120B total (128 experts, 4 active per token)
• Architecture: MoE with sliding + full attention
• Context: 131,072 tokens
• License: Apache 2.0
• Method: Imperceptible noise ghost pass (1e-8 scale)

Available Formats

Format	Size	Use Case
BF16 Safetensors	234GB	Full precision
GGUF Q8_0	116GB	Near-lossless
GGUF Q5_K_M	88GB	Quality sweet spot
GGUF Q4_K_M	82GB	Popular consumer quant
GGUF IQ4_XS	63GB	Maximum compression

About Zero Point AI

Intelligence From The Void. https://zeropointai.uk