NEXUS-Security — MLX bf16

Full-precision (bf16) MLX build of fableforge-ai/NEXUS-Security (Qwen2ForCausalLM, ~1.5B), for fast local inference on Apple Silicon.

• Precision: bf16 — no quantization, identical weights to source.
• Size: ~3 GB · retention: 100% (full precision)
• Converted with: mlx-lm, mlx.

Weights unchanged from the source — format + precision conversion only.

Other builds

GGUF builds (llama.cpp/Ollama): King3Djbl/nexus-security-GGUF.

Usage

pip install mlx-lm

from mlx_lm import load, generate
model, tok = load("ahmedandaloes/NEXUS-Security-MLX-bf16")
msgs=[{"role":"user","content":"Name one common web vulnerability."}]
p=tok.apply_chat_template(msgs, add_generation_prompt=True)
print(generate(model, tok, prompt=p, max_tokens=256, verbose=True))

Attribution & license

• Source: fableforge-ai/NEXUS-Security · base Qwen/Qwen2-1.5B
• License Apache-2.0 (verify against source). MLX build for the Apple Silicon community.
• For authorized security work only.