Neuro-Orchestrator-8B

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the TIES merge method using miromind-ai/MiroThinker-v1.0-8B as a base.

Models Merged

The following models were included in the merge:

• Kwaipilot/HiPO-8B
• nvidia/Nemotron-Orchestrator-8B

Configuration

The following YAML configuration was used to produce this model:

name: Neuro-Orchestrator-8B
merge_method: ties
base_model: miromind-ai/MiroThinker-v1.0-8B
parameters:
  normalize: true
  int8_mask: true
dtype: bfloat16
tokenizer_source: union  # Critical: Merges special tokens from all models

models:
  # Base Model: MiroThinker (The Body/Executioner)
  # We use this as the base so the resulting model defaults to its
  # superior 256k context and tool-use templates.
  - model: miromind-ai/MiroThinker-v1.0-8B
    parameters:
      density: 1.0
      weight: 1.0

  # Model 2: Nvidia Nemotron Orchestrator (The Manager)
  # We want high density (coverage) but lower weight to guide
  # logic without breaking the tool syntax.
  - model: nvidia/Nemotron-Orchestrator-8B
    parameters:
      density: 0.8
      weight: 0.4

  # Model 3: Kwaipilot HiPO (The Controller)
  # We mainly want its 'gating' logic (when to think).
  # A lower density focuses on the most critical policy changes.
  - model: Kwaipilot/HiPO-8B
    parameters:
      density: 0.5
      weight: 0.3