Instructions to use FINAL-Bench/Darwin-4B-Chimera with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use FINAL-Bench/Darwin-4B-Chimera with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="FINAL-Bench/Darwin-4B-Chimera")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("FINAL-Bench/Darwin-4B-Chimera")
model = AutoModelForCausalLM.from_pretrained("FINAL-Bench/Darwin-4B-Chimera")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use FINAL-Bench/Darwin-4B-Chimera with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "FINAL-Bench/Darwin-4B-Chimera"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "FINAL-Bench/Darwin-4B-Chimera",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/FINAL-Bench/Darwin-4B-Chimera

SGLang

How to use FINAL-Bench/Darwin-4B-Chimera with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "FINAL-Bench/Darwin-4B-Chimera" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "FINAL-Bench/Darwin-4B-Chimera",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "FINAL-Bench/Darwin-4B-Chimera" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "FINAL-Bench/Darwin-4B-Chimera",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use FINAL-Bench/Darwin-4B-Chimera with Docker Model Runner:
```
docker model run hf.co/FINAL-Bench/Darwin-4B-Chimera
```

SeaWolf-AI commited on 3 days ago

Commit

91e6efb

verified ·

1 Parent(s): 6c73c4a

Card in English: lead with Chimera value/significance; remove base_model (pointed to private repo -> 401 for visitors)

Browse files

Files changed (1) hide show

README.md +38 -35

README.md CHANGED Viewed

@@ -1,6 +1,5 @@
 ---
 license: gemma
-base_model: FINAL-Bench/Darwin-V9-Chimera-4B-SFT
 tags:
   - chimera
   - darwin
@@ -17,49 +16,53 @@ library_name: transformers
 # Darwin-4B-Chimera
-**비드래프트(VIDRAFT) Chimera 기술로 만든 4B 한국어 추론 모델.**
-작은 모델이 자기 크기를 넘어서게 만드는 것이 목표입니다. 파라미터를 늘리는 대신, 서로 다른 강점을 **합체**하고 스스로 **정제**하게 했습니다.
 ---
-## Chimera란
-**Chimera**는 서로 다른 계열·강점을 가진 모델의 구성요소를 **하나의 모델로 합체**시키는 비드래프트의 모델 진화 기술입니다.
-- 각 부모의 지식을 **손상 없이 보존**한 채 합칩니다.
-- 단일 부모로는 도달할 수 없는 조합을 만듭니다 — 결과물은 어느 한 부모로도 환원되지 않습니다.
-- **추가 사전학습 없이** 세대를 이어갑니다.
-이 모델은 Chimera 4B 계보의 검증된 개체에 비드래프트 자체 **RFT(강화 미세조정)** 파이프라인을 적용해, 한국어 지식 추론을 끌어올린 결과물입니다.
----
-## 성능
-**KMMLU 6과목 · 240문항 held-out · greedy 디코딩**
-| 모델 | KMMLU | vs 기준선 |
-|---|---|---|
-| 기준선 (Chimera 4B) | 42.9% | — |
-| **Darwin-4B-Chimera (본 모델)** | **48.3%** | **+5.4%p** |
-같은 가중치·같은 하네스에서 정제 단계만 달리해 측정했습니다. 6과목 중 다수에서 상승했습니다.
 ---
-## 정직한 범위 (Honest scope)
-수치를 부풀리지 않기 위해 한계를 먼저 밝힙니다.
-- **평가 셋이 작습니다.** 240문항 기준 표준오차 약 ±1.4%p입니다. 더 큰 셋에서의 재확인이 필요합니다.
-- **KMMLU 6과목 한정** 수치이며, 전 도메인 일반화를 뜻하지 않습니다.
-- **4B 규모** 모델입니다. 플래그십(Darwin-398B 등)과 직접 비교 대상이 아닙니다.
-- **greedy 디코딩** 기준입니다. 디코딩·프롬프트가 바뀌면 결과가 달라질 수 있습니다.
 ---
-## 사용
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
@@ -76,22 +79,22 @@ print(tok.decode(out[0][ids.shape[-1]:], skip_special_tokens=True))
 ---
-## 방법론 공개 범위
-Chimera 합체 구성과 RFT 정제의 내부 설계(데이터 선별 기준·학습 설정·라우팅 방식 등)는 **비드래프트 영업기밀**입니다.
-> **결과는 공개하고, 레시피는 공개하지 않습니다.**
-모델 가중치와 측정 수치는 공개되어 있으므로, 누구나 위 결과를 직접 재현·검증할 수 있습니다.
 ---
-## 관련 링크
-- **Darwin 패밀리** — https://huggingface.co/collections/FINAL-Bench/darwin-family
-- **진화적 병합 방법론 논문** — arXiv [2605.14386](https://arxiv.org/abs/2605.14386)
-- **비드래프트** — https://vidraft.net
-## 라이선스
-`gemma` — 합체 계보에 Gemma 파생 구성요소가 포함되어 Gemma 라이선스를 따릅니다. 사용 전 [Gemma Terms of Use](https://ai.google.dev/gemma/terms)를 확인하십시오.

 ---
 license: gemma
 tags:
   - chimera
   - darwin
 # Darwin-4B-Chimera
+**A 4B Korean-reasoning model built with VIDRAFT's Chimera technology.**
+Most models get better by getting bigger. This one got better by **combining** — and then refining itself.
 ---
+## Why Chimera matters
+Merging models usually means losing something. Average two networks and you get a compromise: each parent's edge is blunted, and the child is worse than both at what each did best. That is why "model merging" has a reputation as a cheap trick.
+Chimera is built on the opposite premise.
+### 1. The parents survive intact
+Chimera fuses components from models of **different families and different strengths** while preserving what each parent knew. Knowledge is carried over, not averaged away. What you merge in is what you keep.
+### 2. The result cannot be reduced to any single parent
+A Chimera individual is not "model A with a bit of B." It sits at a point in model space that **no parent reaches alone**, and it cannot be recovered from any one of them. That non-reducibility is the entire point — it is what makes a lineage an asset rather than a re-skin.
+### 3. No additional pretraining
+A new generation is a **fusion plus an evaluation**, not a training run measured in GPU-months. Iteration collapses from months to days — which means the space of viable combinations actually gets explored instead of theorized about.
+### 4. It compounds
+Chimera individuals become parents. Strengths accumulate across generations: parent → child → grandchild. Capability is **grown**, not purchased.
+### 5. It answers the real constraint
+Frontier capability is gated by capital: tens of thousands of GPUs, months of training, hundreds of millions of dollars. Chimera attacks that gate directly — **structural evolution instead of brute-force scale**. It is how a small team competes on method rather than on budget.
 ---
+## What that buys you at 4B
+This model is a late-generation Chimera individual, refined further by VIDRAFT's own reinforcement fine-tuning pipeline.
+**KMMLU · 6 subjects · 240 held-out items · greedy decoding**
+| Model | KMMLU | Δ |
+|---|---|---|
+| Chimera 4B baseline | 42.9% | — |
+| **Darwin-4B-Chimera** | **48.3%** | **+5.4pp** |
+**+5.4pp on Korean knowledge reasoning with zero parameter growth** — 4.02B before, 4.02B after. The gain came from what the model learned from itself, not from more weights. Same architecture, same size, same inference cost: strictly better.
+That matters because 4B is the size that actually ships. It runs on a single consumer GPU, on-premise, inside an air-gapped network — the places where frontier APIs cannot go.
 ---
+## Usage
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
 ---
+## What is open, and what is not
+The **weights are open** and the **numbers are stated with the exact evaluation setup**, so anyone can reproduce the result above rather than take our word for it.
+The internal design of Chimera fusion and of the refinement pipeline — component selection, data selection criteria, training configuration, routing — is **VIDRAFT proprietary**.
+> **Results are open. The recipe is not.**
 ---
+## Links
+- **Darwin family** — https://huggingface.co/collections/FINAL-Bench/darwin-family
+- **Method paper (evolutionary merging)** — arXiv [2605.14386](https://arxiv.org/abs/2605.14386)
+- **VIDRAFT** — https://vidraft.net
+## License
+`gemma` — the Chimera lineage includes Gemma-derived components, so the Gemma license applies. Review the [Gemma Terms of Use](https://ai.google.dev/gemma/terms) before use.