Instructions to use Velvet-Capital/VelvetFlash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Velvet-Capital/VelvetFlash with Transformers:

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

pipe = pipeline("text-generation", model="Velvet-Capital/VelvetFlash")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Velvet-Capital/VelvetFlash")
model = AutoModelForCausalLM.from_pretrained("Velvet-Capital/VelvetFlash", device_map="auto")
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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Velvet-Capital/VelvetFlash with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Velvet-Capital/VelvetFlash

SGLang

How to use Velvet-Capital/VelvetFlash 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 "Velvet-Capital/VelvetFlash" \
    --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": "Velvet-Capital/VelvetFlash",
		"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 "Velvet-Capital/VelvetFlash" \
        --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": "Velvet-Capital/VelvetFlash",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Velvet-Capital/VelvetFlash with Docker Model Runner:
```
docker model run hf.co/Velvet-Capital/VelvetFlash
```

Velvet-Capital commited on 17 days ago

Commit

ac9add1

verified ·

1 Parent(s): 52c1b00

Rename to Velvet Flash 4B

Browse files

Files changed (5) hide show

README.md +7 -7
assets/comp_efficiency.png +0 -0
assets/comp_leaderboard.png +0 -0
assets/heatmap.png +0 -0
assets/per_dim.png +0 -0

README.md CHANGED Viewed

@@ -7,9 +7,9 @@ language: [en, zh]
 tags: [crypto, defi, trading, agent, tool-routing, function-calling, qwen3, velvet]
 ---
-# Velvet Flash 4
-**Velvet Flash 4** is a compact 4B-parameter crypto **action-routing** model by [Velvet Capital](https://velvet.capital), built on [Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B). On the crypto-skill-benchmark it **ranks #1 against open-source models up to 17× its size** (Qwen3.5-27B, DeepSeek-V4, Llama-3.3-70B, Gemma-4-31B, Mistral-24B).
 You give it a crypto skill (an exchange API, a DEX, a wallet SDK or a DeFi protocol, described in a `SKILL.md`-style system prompt) and a request in plain language. It turns the request into the **exact command or API call** for that skill. It also writes a **human-readable confirmation summary** and adds **safety warnings** when a request looks risky.
@@ -139,7 +139,7 @@ Reply **confirm** to execute, or tell me what to change.
 **vLLM (OpenAI-compatible server):**
 ```bash
-vllm serve Velvet-Capital/VelvetFlash --served-model-name velvet-flash-4 --max-model-len 16384
 # send "chat_template_kwargs": {"enable_thinking": false} with each request
 ```
@@ -167,13 +167,13 @@ Evaluated on [crypto-skill-benchmark](https://github.com/Minara-AI/crypto-skill-
 ### Comparison with open-source models
-Velvet Flash 4 **ranks #1 of 7**. It outscores open-source generalists that are **6–17× larger**.
 ![Overall leaderboard](assets/comp_leaderboard.png)
 | Model | Size | Overall | Safety | Coverage | Robustness | Routing | UX |
 |---|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
-| **★ Velvet Flash 4** | **4B** | **50** | **61** | **34** | 53 | **47** | **52** |
 | Qwen3.5-27B | 27B | 45 | 46 | 25 | **71** | 39 | 48 |
 | DeepSeek-V4-Flash | MoE | 40 | 39 | 29 | 54 | 39 | 41 |
 | Llama-3.3-70B | 70B | 32 | 33 | 22 | 44 | 32 | 30 |
@@ -187,7 +187,7 @@ Velvet Flash 4 **ranks #1 of 7**. It outscores open-source generalists that are
 | Model | Minara | Binance Spot | OKX DEX | Uniswap | GMX | MetaMask |
 |---|:--:|:--:|:--:|:--:|:--:|:--:|
-| **★ Velvet Flash 4** | 66 | 34 | **50** | 47 | **47** | **54** |
 | Qwen3.5-27B | 70 | 44 | 45 | **48** | 34 | 27 |
 | DeepSeek-V4-Flash | 44 | **45** | 47 | 40 | 37 | 26 |
 | Llama-3.3-70B | **75** | 32 | 26 | 19 | 18 | 21 |
@@ -198,7 +198,7 @@ Velvet Flash 4 **ranks #1 of 7**. It outscores open-source generalists that are
 ![Full results matrix](assets/heatmap.png)
 **Key takeaways**
-- **Safety leader:** Velvet Flash 4 scores highest on safety, the dimension with the most weight (61, against 46 for the next best). It is the most reliable model here at confirming before moving funds and at parsing amounts exactly.
 - **Strongest on the hardest platforms:** it leads on OKX DEX, GMX and MetaMask, where the larger models struggle. MetaMask is 54 against 27 for the next best.
 - **Most consistent:** some larger models beat it on individual platforms (Llama-3.3-70B and Qwen3.5-27B on Minara, DeepSeek and Qwen on Binance Spot). None of them matches its consistency across all six, which is what the weighted average rewards.
 - **More than double its base model:** it scores 50 against 23 for Qwen3-4B and improves on every platform and every dimension.

 tags: [crypto, defi, trading, agent, tool-routing, function-calling, qwen3, velvet]
 ---
+# Velvet Flash 4B
+**Velvet Flash 4B** is a compact 4B-parameter crypto **action-routing** model by [Velvet Capital](https://velvet.capital), built on [Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B). On the crypto-skill-benchmark it **ranks #1 against open-source models up to 17× its size** (Qwen3.5-27B, DeepSeek-V4, Llama-3.3-70B, Gemma-4-31B, Mistral-24B).
 You give it a crypto skill (an exchange API, a DEX, a wallet SDK or a DeFi protocol, described in a `SKILL.md`-style system prompt) and a request in plain language. It turns the request into the **exact command or API call** for that skill. It also writes a **human-readable confirmation summary** and adds **safety warnings** when a request looks risky.
 **vLLM (OpenAI-compatible server):**
 ```bash
+vllm serve Velvet-Capital/VelvetFlash --served-model-name velvet-flash-4b --max-model-len 16384
 # send "chat_template_kwargs": {"enable_thinking": false} with each request
 ```
 ### Comparison with open-source models
+Velvet Flash 4B **ranks #1 of 7**. It outscores open-source generalists that are **6–17× larger**.
 ![Overall leaderboard](assets/comp_leaderboard.png)
 | Model | Size | Overall | Safety | Coverage | Robustness | Routing | UX |
 |---|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
+| **★ Velvet Flash 4B** | **4B** | **50** | **61** | **34** | 53 | **47** | **52** |
 | Qwen3.5-27B | 27B | 45 | 46 | 25 | **71** | 39 | 48 |
 | DeepSeek-V4-Flash | MoE | 40 | 39 | 29 | 54 | 39 | 41 |
 | Llama-3.3-70B | 70B | 32 | 33 | 22 | 44 | 32 | 30 |
 | Model | Minara | Binance Spot | OKX DEX | Uniswap | GMX | MetaMask |
 |---|:--:|:--:|:--:|:--:|:--:|:--:|
+| **★ Velvet Flash 4B** | 66 | 34 | **50** | 47 | **47** | **54** |
 | Qwen3.5-27B | 70 | 44 | 45 | **48** | 34 | 27 |
 | DeepSeek-V4-Flash | 44 | **45** | 47 | 40 | 37 | 26 |
 | Llama-3.3-70B | **75** | 32 | 26 | 19 | 18 | 21 |
 ![Full results matrix](assets/heatmap.png)
 **Key takeaways**
+- **Safety leader:** Velvet Flash 4B scores highest on safety, the dimension with the most weight (61, against 46 for the next best). It is the most reliable model here at confirming before moving funds and at parsing amounts exactly.
 - **Strongest on the hardest platforms:** it leads on OKX DEX, GMX and MetaMask, where the larger models struggle. MetaMask is 54 against 27 for the next best.
 - **Most consistent:** some larger models beat it on individual platforms (Llama-3.3-70B and Qwen3.5-27B on Minara, DeepSeek and Qwen on Binance Spot). None of them matches its consistency across all six, which is what the weighted average rewards.
 - **More than double its base model:** it scores 50 against 23 for Qwen3-4B and improves on every platform and every dimension.

assets/comp_efficiency.png CHANGED Viewed

assets/comp_leaderboard.png CHANGED Viewed

assets/heatmap.png CHANGED Viewed

assets/per_dim.png CHANGED Viewed