Jarvis-7B-Unified-Coder

Jarvis-7B-Unified-Coder is a high-performance genetic sibling fusion anchored on Qwen/Qwen2.5-Coder-7B-Instruct. It combines four specialized models fine-tuned on the identical Coder-7B architecture using TIES sign-consensus:

  1. Abliterated Zero-Refusal Foundation (35%): Goekdeniz-Guelmez/Josiefied-Qwen2.5-Coder-7B-Instruct-abliterated-v1
    • Strips preachy alignment vectors; directly answers cybersecurity and engineering queries without canned refusals.
  2. Microsoft SOTA Code Editing (30%): microsoft/NextCoder-7B
    • ICML 2025 SOTA code-editing model for deep multi-turn code refactoring and syntax integrity.
  3. Outflank Red-Team Specialist (20%): outflanknl/Dante-7B
    • Specialized in low-level systems programming, loaders, and offensive security analysis.
  4. Jarvis Autonomous CyberAgent (15%): adityaxess/Jarvis-7B-CyberAgent-Adapter
    • Autonomous security incident triage, vulnerability detection, and structured JSON tool-calling.

Merge Details

  • Method: TIES Consensus
  • Base Anchor: Qwen/Qwen2.5-Coder-7B-Instruct
  • Dtype: float16
  • Density: 0.65
  • Format: Native Safetensors + ChatML Tokenizer
Downloads last month
22
Safetensors
Model size
8B params
Tensor type
F16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for adityaxess/Jarvis-7B-Unified-Coder