---
license: apache-2.0
base_model:
- Naphula/BeaverAI_Fallen-Mistral-Small-3.1-24B-v1e_textonly
- TheDrummer/Cydonia-24B-v4.3
- TheDrummer/Magidonia-24B-v4.3
- TheDrummer/Precog-24B-v1
- TheDrummer/Rivermind-24B-v1
- Darkhn/Magistral-2509-24B-Text-Only
language:
- en
library_name: transformers
tags:
- mergekit
- merge
- mistral
widget:
  - text: "Mergedonia-PROMETHEUS-24B-v1"
    output:
      url: https://cdn-uploads.huggingface.co/production/uploads/68e840caa318194c44ec2a04/ZLfHoqVelJKdSxxY2Tst5.jpeg
---

<pre><span style="color: #FF4500;">PROMETHEUS (VERSION P): STABILIZED</span></pre>

![prometheus](https://cdn-uploads.huggingface.co/production/uploads/68e840caa318194c44ec2a04/ZLfHoqVelJKdSxxY2Tst5.jpeg)

# 🌄 Mergedonia PROMETHEUS 24B v1

A highly experimental new merge method which combines several official and custom methods into a complex 6-stage pipeline.

```yaml
architecture: MistralForCausalLM
merge_method: prometheus # Version P
base_model: B:/24B/Darkhn--Magistral-2509-24B-Text-Only
parameters:
  # --- CABS + RAM Core Settings ---
  default_n_m_ratio: "1/4" 
  ram_r: 0.1
  ram_alpha: 0.2
  della_eps: 0.05
  gamma: 0.001
  breadcrumbs_density: 0.9
  select_topk: 0.5     
  # --- Advanced Toggles ---
  use_sign_consensus: true 
  della_row_wise: true
  sce_weighting: true
  lambda: 1.0
  auto_pad: true
  skip_on_mismatch: true
  # --- Quantum Frankenstein Glue + Shannon Entropy ---
  iota: 0.1
  zeta: 16
  chi: 0.15
  gamma: 1.0
  # --- AETHERIC SYMPLECTIC GROUNDING ---
  asg_resonance: 0.3
  asg_grounding: 0.7
  # --- FLUX Frankenstein Surgery (QUICK TEST PLATINUM) ---
  frankenstein_surgery: true
  phi_optim: true
  phi: 0.4
  rho: 0.15
  mu: 0.5
  tau: 2.5
  alpha: 0.5
  # --- Hilbert Transform Slingshot ---
  eta: 1.2              
  delta: 0.915
  mercury_retrograde_gravity: 1.0
  pi: 3141
  tol: 1.0e-9
  max_iter: 100
  kappa: 0.8
models:
  # --- TIER 1 ---
  - model: B:/24B/!models--TheDrummer--Precog-24B-v1
    parameters:
      weight: 1.0
      n_m_ratio: "1/5" ## Balanced for Mergedonia ## "1/2" # 50% density. Gets first pick and claims a massive footprint.
      allow_sce_steal: true
      fill_remaining: false
  # --- TIER 2 ---
  - model: B:/24B/TheDrummer--Magidonia-24B-v4.3
    parameters:
      weight: 1.0
      n_m_ratio: "1/4" ## 25% density. Claims a fourth of whatever Precog left behind. ##
      allow_sce_steal: true
      fill_remaining: false
  # --- TIER 3 ---
  - model: B:/24B/TheDrummer--Cydonia-24B-v4.3
    parameters:
      weight: 1.0
      n_m_ratio: "1/3" # No override: Falls back to global default[1, 4] (25% density)
      allow_sce_steal: true
      fill_remaining: false
  - model: B:/24B/!BeaverAI_Fallen-Mistral-Small-3.1-24B-v1e_textonly
    parameters:
      weight: 1.0
      n_m_ratio: "1/2" # No override: Falls back to global default [1, 4] (25% density)
      allow_sce_steal: true
      fill_remaining: false
  - model: B:/24B/!models--TheDrummer--Rivermind-24B-v1
    parameters:
      weight: 1.0
      n_m_ratio: "1/2" ## We use an increasing n_m_ratio for balanced Mergedonia, but normally a decreasing ratio produces the desired CABS model "hierarchy" ##
      # No override: Falls back to global default [1, 4] (25% density)
      allow_sce_steal: true # false # STOP THE STEAL: Rivermind can only have leftovers
      fill_remaining: true
tokenizer:
  source: union
chat_template: auto
dtype: float32
out_dtype: bfloat16
name: 🌄 Mergedonia-Prometheus-24B-v1
```