---
license: apache-2.0
language:
- en
- zh
- ko
- ja
- multilingual
library_name: transformers
pipeline_tag: text-generation
tags:
- darwin
- darwin-v7
- evolutionary-merge
- merge
- mergekit
- reasoning
- advanced-reasoning
- chain-of-thought
- thinking
- qwen3.6
- qwen
- claude-opus
- distillation
- multilingual
- gpqa
- benchmark
- open-source
- apache-2.0
- hybrid-vigor
- proto-agi
- vidraft
- eval-results
base_model:
- Qwen/Qwen3.6-27B
- rico03/Qwen3.6-27B-Claude-Opus-Reasoning-Distilled
base_model_relation: merge
model-index:
- name: Darwin-28B-Opus
results:
- task:
type: text-generation
name: Graduate-Level Reasoning
dataset:
type: Idavidrein/gpqa
name: GPQA Diamond
config: gpqa_diamond
split: train
metrics:
- type: accuracy
value: 88.89
name: Accuracy
verified: false
---
# Darwin-28B-Opus — Qwen3.6-27B × Opus-Distilled Evolutionary Merge
-blue?style=for-the-badge)
-purple?style=for-the-badge)
-blue?style=for-the-badge)
-blue?style=for-the-badge)
-blue?style=for-the-badge)
> Qwen3.6-27B dense · 27.6B parameters · Hybrid Linear/Full Attention · BF16 · Thinking Mode · Apache 2.0
> **Darwin V7 evolutionary merge: Father × Opus-distilled Mother → 88.89% on GPQA Diamond (3-stage adaptive evaluation)**
---
## Abstract
**Darwin-28B-Opus** is the first reasoning model of the Darwin series built on the **Qwen3.6 generation** backbone. Produced by the Darwin V7 evolutionary breeding engine from two publicly available parents, it combines the strong bilingual reasoning of Qwen3.6-27B with Claude Opus 4-style chain-of-thought distilled behaviour.
On the **GPQA Diamond** graduate-level reasoning benchmark (198 PhD-level questions), Darwin-28B-Opus scores **88.89 %** under the standard 3-stage adaptive evaluation, slightly edging out its larger MoE sibling Darwin-36B-Opus (88.4 %) and clearly surpassing its Qwen3.5-generation counterpart Darwin-27B-Opus (86.9 %).
---
## 🧬 Model Lineage
| Role | Model | Role in the Merge |
|:---:|:---|:---|
| **Father (父)** | [`Qwen/Qwen3.6-27B`](https://huggingface.co/Qwen/Qwen3.6-27B) | Qwen3.6 generation dense backbone with hybrid linear/full attention. |
| **Mother (母)** | [`rico03/Qwen3.6-27B-Claude-Opus-Reasoning-Distilled`](https://huggingface.co/rico03/Qwen3.6-27B-Claude-Opus-Reasoning-Distilled) | Claude Opus reasoning-distilled variant of the same backbone (Jackrong-style distillation, 14 k traces). |
| **Offspring** | **`Darwin-28B-Opus`** (this model) | Darwin V7 evolutionary merge; Qwen3.6 architecture retained, Opus reasoning style inherited. |
> **Why 28B?** The `28B` label denotes the Qwen3.6-generation member of the Darwin lineup (`+1` over the Qwen3.5-era `Darwin-27B-Opus`).
> The actual parameter count is **27.6 B**, and the architecture exactly follows Qwen3.6-27B.
---
## ⚙️ Technical Specifications
| Component | Value |
|:---|:---|
| Architecture | `Qwen3_5ForConditionalGeneration` (Qwen3.6 generation, hybrid linear + full attention) |
| Parameters | **27.6 B** (BF16) |
| Hidden size | 5 120 |
| Intermediate size | 17 408 |
| Head dim | 256 |
| Layers | 64 (3 linear : 1 full attention, `full_attention_interval = 4`) |
| Precision | bfloat16 |
| Context length | Inherited from base (long-chain reasoning supported) |
| License | Apache 2.0 |
---
## 🏆 Benchmark — GPQA Diamond (198 questions)
Darwin-28B-Opus is evaluated under our standard **3-stage adaptive evaluation** protocol, identical to the protocol used across the Darwin series.
| Stage | Decoding Protocol | Cost | **Accuracy** |
|:---:|:---|:---:|:---:|
| **Stage 1** | Single-shot greedy baseline | 1× | **74.75 %** (148 / 198) |
| **Stage 2** | Majority vote ×8 at temperature 0.7 on Stage-1 wrongs | 8× | **83.84 %** (166 / 198) |
| **Stage 3** | Adaptive ensemble refinement (close-tie tiebreaker + iterative MTI on residual hard questions) | ≈ 20× | **🥇 88.89 %** (176 / 198) |
**Key performance indicators**:
- Stage 1 → Stage 3: **+14.14 %p** through adaptive protocol
- vs Darwin-27B-Opus (86.9 %): **+1.99 %p**
- vs Darwin-36B-Opus (88.4 %): **+0.49 %p**
- vs Darwin-31B-Opus (85.9 %): **+2.99 %p**
---
## 🚀 Usage
### Standard inference (Stage 1 baseline)
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
tok = AutoTokenizer.from_pretrained(
"FINAL-Bench/Darwin-28B-Opus",
trust_remote_code=True,
)
model = AutoModelForCausalLM.from_pretrained(
"FINAL-Bench/Darwin-28B-Opus",
torch_dtype=torch.bfloat16,
device_map="auto",
trust_remote_code=True,
)
messages = [
{"role": "user",
"content": "Solve: If f(x) = x³ − 3x + 2, find all critical points and classify them."}
]
text = tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tok(text, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=2048, do_sample=False)
print(tok.decode(outputs[0][inputs.input_ids.shape[-1]:], skip_special_tokens=True))
```
### Enhanced accuracy (Stage 2-3 adaptive)
For leaderboard-grade accuracy, combine:
1. Stage 1 greedy baseline,
2. Stage 2 maj@8 temperature sampling on low-confidence answers,
3. Stage 3 adaptive refinement on still-disputed answers.
Reference implementation is provided in the Darwin-series evaluation harness.
---
## 🎯 Recommended Use-Cases
- **Graduate-level STEM reasoning** (GPQA / science qualifying exams)
- **Mathematical problem solving** (MATH, AIME-style problems)
- **Code generation and debugging** (HumanEval, MBPP)
- **Complex multi-step chain-of-thought tasks**
- **Bilingual reasoning** (strong English + Korean; also Chinese / Japanese)
## ⚠️ Limitations
- At 27.6 B parameters in bfloat16, full inference requires ≈ 55 GB of VRAM (e.g., a single A100-80GB or B200).
- Optimised for English first, with secondary support for Korean, Chinese, and Japanese.
- Deep Opus-style reasoning traces tend to be verbose — control with `max_new_tokens` as needed.
---
## 📚 Citation
```bibtex
@misc{darwin28b_opus_2026,
title = {Darwin-28B-Opus: Evolutionary Merging of Qwen3.6-27B with Claude-Opus-Distilled Reasoning},
author = {FINAL-Bench / Darwin Research Team},
year = {2026},
howpublished = {\url{https://huggingface.co/FINAL-Bench/Darwin-28B-Opus}},
note = {Darwin V7 · Mother-centric Ratio Interpolation merge · 88.89 % GPQA Diamond (3-stage)}
}
```
---
## 🔗 Related Darwin Models
- **Darwin-36B-Opus** — MoE 36B, Qwen3.6-35B-A3B × Opus distilled, GPQA 88.4 %
- **Darwin-31B-Opus** — 31B dense, multilingual-strong reasoning, GPQA 85.9 %
- **Darwin-27B-Opus** — 27B dense (Qwen3.5 generation), GPQA 86.9 %
- **Darwin-9B-NEG** — 9B with Native Entropy Gating, GPQA 84.3 %
- **Darwin-9B-Opus** — the Qwen3.5-9B Darwin member
- **Darwin-4B-Genesis** — smallest Darwin member
---
*Darwin V7 · Qwen3.6 generation flagship · Sealed 2026-04-25 · FINAL-Bench*