Lumix series — agentic coding models and formats.
Sir Frosty PRO
Blackfrost-AI
AI & ML interests
Self-hosted, security-first AI. Agentic CLI harnesses and multi-agent operator systems; fine-tuning open models (LoRA/PEFT) into domain-specialized agents; offensive + defensive AI security and model robustness. Most drawn to the orchestration layer where autonomous agents meet real infrastructure and have to actually ship work and to running capable models on hardware you own.
Recent Activity
updated a model about 1 hour ago
Blackfrost-AI/Qwen3.8-Flash-Next-DERISKED-NVFP4 published a model about 1 hour ago
Blackfrost-AI/Qwen3.8-Flash-Next-DERISKED-NVFP4 updated a model about 3 hours ago
Blackfrost-Research/GLM-5.3-Flash-DERISKED-BF16Organizations
THE QWEN FAMILY FINE-TUNED MODELS IN THE VOID
-
Blackfrost-AI/PINQWEN-3.6-27B-NVFP4-ABLITERATED
Text Generation • 17B • Updated • 187 • 2 -
Blackfrost-AI/PINQWEN-3.6-35B-CLEAN-BF16
Text Generation • 36B • Updated • 71 • 1 -
Blackfrost-AI/PINQWEN-3.5-9B-1M-GGUF
Text Generation • 9B • Updated • 1.39k • 2 -
Blackfrost-AI/PINQWEN-3.5-9B-1M-NVFP4
Text Generation • 7B • Updated • 123 • 2
PINQWEN series — long-context reasoning models and quants.
-
Blackfrost-AI/PINQWEN-3.5-9B-1M-BF16
Text Generation • 10B • Updated • 96 • 2 -
Blackfrost-AI/PINQWEN-3.5-9B-1M-NVFP4
Text Generation • 7B • Updated • 123 • 2 -
Blackfrost-AI/PINQWEN-3.5-9B-1M-GGUF
Text Generation • 9B • Updated • 1.39k • 2 -
Blackfrost-AI/PINQWEN-3.6-27B-NVFP4-ABLITERATED
Text Generation • 17B • Updated • 187 • 2
versions of glm 5.2 made for security research 0 refusals. It will ask for assurance but never refuses
Lumix series — agentic coding models and formats.
PINQWEN series — long-context reasoning models and quants.
-
Blackfrost-AI/PINQWEN-3.5-9B-1M-BF16
Text Generation • 10B • Updated • 96 • 2 -
Blackfrost-AI/PINQWEN-3.5-9B-1M-NVFP4
Text Generation • 7B • Updated • 123 • 2 -
Blackfrost-AI/PINQWEN-3.5-9B-1M-GGUF
Text Generation • 9B • Updated • 1.39k • 2 -
Blackfrost-AI/PINQWEN-3.6-27B-NVFP4-ABLITERATED
Text Generation • 17B • Updated • 187 • 2
THE QWEN FAMILY FINE-TUNED MODELS IN THE VOID
-
Blackfrost-AI/PINQWEN-3.6-27B-NVFP4-ABLITERATED
Text Generation • 17B • Updated • 187 • 2 -
Blackfrost-AI/PINQWEN-3.6-35B-CLEAN-BF16
Text Generation • 36B • Updated • 71 • 1 -
Blackfrost-AI/PINQWEN-3.5-9B-1M-GGUF
Text Generation • 9B • Updated • 1.39k • 2 -
Blackfrost-AI/PINQWEN-3.5-9B-1M-NVFP4
Text Generation • 7B • Updated • 123 • 2
versions of glm 5.2 made for security research 0 refusals. It will ask for assurance but never refuses