Overview

Darwin-31B-Opus is a reasoning-enhanced model created by merging google/gemma-4-31B-it (Father) and TeichAI/gemma-4-31B-it-Claude-Opus-Distill (Mother) using the Darwin V6 engine.

Darwin V6 diagnoses both parent models at the tensor level before merging, assigning an independent optimal ratio to each of the 1,188 tensors. This is fundamentally different from conventional merging tools that apply a single uniform ratio across all tensors.

---

Parent Models

Model Diagnostic Scan (MDS)

Left: Father (gemma-4-31B-it) — balanced generalist with low activation across most probes. Right: Mother (Claude-Opus-Distill) — strong REASONING concentration in L50-L60, CODE activation in late layers, KOREAN at start and end. The Mother shows significantly more specialized layer patterns from Claude Opus distillation.

---

🏆 Benchmark — GPQA Diamond (198 questions)

GPQA Diamond is a 198-question, PhD-level graduate science reasoning benchmark.

The 85.9 % GPQA Diamond result is produced with the Darwin-DELPHI test-time reasoning engine applied on top of this model. The evaluation methodology is protected; sample counts, staging, and thresholds are a trade secret. ARC-Challenge 82.89 % is the internal evolutionary-selection score used during the Darwin V6 merge search.

Note: the Gemma 4 architecture (Gemma4ForConditionalGeneration) has a multimodal wrapper that limits lm-eval loglikelihood compatibility; generative evaluation is the valid path for Gemma 4 based models, and Darwin-DELPHI evaluates generatively accordingly.

---

Darwin V6 vs Conventional Merging

---

How Darwin V6 Works

Darwin V6 does not use mergekit or any external merge library. It re-implements DARE-TIES (Yadav et al., 2023) directly via PyTorch tensor operations with per-tensor diagnostic ratios.

Before merging, Darwin performs a Model Diagnostic Scan (MDS) on both parents. For every tensor, it measures Shannon entropy (information density), standard deviation (activation spread), and L2 norm (energy). Additionally, 5 diagnostic probes (REASONING, CODE, MATH, KNOWLEDGE, LANGUAGE) are passed through the model, measuring cosine distance when each layer is skipped to determine functional importance.

The final merge ratio for each tensor:

markdown
Copy code
static_score = entropy × 0.3 + std × 0.2 + clamp(norm, 100) × 0.002
probe_score  = Σ(cosine_distance[probe_i] × weight_i)
combined     = static × 0.4 + probe × 0.6
mri_ratio    = combined_b / (combined_a + combined_b)
final_ratio  = mri_ratio × mri_trust + genome_ratio × (1 - mri_trust)

The mri_trust parameter itself is optimized by the CMA-ES evolutionary algorithm, allowing the system to automatically determine the optimal balance between diagnostic prescription and evolutionary search for each model pair.

After merging, a Health Check compares the child model against both parents layer-by-layer, detecting interference (child importance >> parent max) or function loss (parent importance high but child dropped).

Parent Comparison (MDS Result)

---

Evolution Result

Optimal Genome (14-dimensional adaptive):

markdown
Copy code
global_ratio:        0.5147   (overall merge ratio)
attn_ratio:          0.3169   (Attention layers — Father dominant)
ffn_ratio:           0.9316   (FFN layers — Mother dominant)
embed_ratio:         0.7748   (Embedding)
density_a:           0.8997   (Father DARE density)
density_b:           0.9539   (Mother DARE density)
block_0_ratio:       0.6628   (L0-L9)
block_1_ratio:       0.6431   (L10-L19)
block_2_ratio:       0.5146   (L20-L29, balanced)
block_3_ratio:       0.5971   (L30-L39)
block_4_ratio:       0.6339   (L40-L49)
block_5_ratio:       0.8583   (L50-L59, reasoning core — Mother dominant)
mri_trust:           0.3631   (MDS 36% + Genome 64%)
merge_method_weight: 0.6897

Key observations from the genome: ffn_ratio=0.93 indicates the FFN layers strongly favor the Mother (Claude Opus Distill), and block_5 (L50-L59)=0.86 shows the reasoning core layers also favor Mother. This aligns with the MDS heatmap pattern where Mother's reasoning capability concentrated in the final layers. Meanwhile, attn_ratio=0.32 preserves Father's attention structure, maintaining the original Gemma 4 multimodal and long-context capabilities.

---

Model Specifications

---

Usage

Transformers

python
Copy code
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

tokenizer = AutoTokenizer.from_pretrained("FINAL-Bench/Darwin-31B-Opus", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    "FINAL-Bench/Darwin-31B-Opus",
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True,
)

messages = [{"role": "user", "content": "Prove that sqrt(2) is irrational."}]
text = tokenizer.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True, enable_thinking=True
)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=4096, do_sample=False)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))

---

VRAM Requirements

---

References

• DARE-TIES: Yadav et al., 2023 (https://arxiv.org/abs/2311.03099) — re-implemented, not library-dependent
• Darwin V6 Engine: https://huggingface.co/spaces/ginigen-ai/DARWIN-V5-BACKUP
• FINAL Bench: https://huggingface.co/spaces/FINAL-Bench/Leaderboard

---

Built By

---

Citation

bibtex
Copy code
@misc{vidraft_darwin_31b_opus,
  title        = {Darwin-31B-Opus: Diagnostic-Guided Evolutionary Merge on Gemma 4},
  author       = {VIDRAFT},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/FINAL-Bench/Darwin-31B-Opus}}
}

This model is introduced in Darwin Family.