Instructions to use deepbot-studio/deepbot-reflex-0.8b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepbot-studio/deepbot-reflex-0.8b with Transformers:

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

pipe = pipeline("text-generation", model="deepbot-studio/deepbot-reflex-0.8b")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("deepbot-studio/deepbot-reflex-0.8b")
model = AutoModelForMultimodalLM.from_pretrained("deepbot-studio/deepbot-reflex-0.8b", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use deepbot-studio/deepbot-reflex-0.8b with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Use Docker

docker model run hf.co/deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use deepbot-studio/deepbot-reflex-0.8b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deepbot-studio/deepbot-reflex-0.8b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "deepbot-studio/deepbot-reflex-0.8b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

SGLang

How to use deepbot-studio/deepbot-reflex-0.8b 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 "deepbot-studio/deepbot-reflex-0.8b" \
    --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": "deepbot-studio/deepbot-reflex-0.8b",
		"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 "deepbot-studio/deepbot-reflex-0.8b" \
        --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": "deepbot-studio/deepbot-reflex-0.8b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use deepbot-studio/deepbot-reflex-0.8b with Ollama:
```
ollama run hf.co/deepbot-studio/deepbot-reflex-0.8b:Q4_K_M
```
Unsloth Desktop

How to use deepbot-studio/deepbot-reflex-0.8b with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "deepbot-studio/deepbot-reflex-0.8b:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use deepbot-studio/deepbot-reflex-0.8b with Docker Model Runner:
```
docker model run hf.co/deepbot-studio/deepbot-reflex-0.8b:Q4_K_M
```

Lemonade

How to use deepbot-studio/deepbot-reflex-0.8b with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Run and chat with the model

lemonade run user.deepbot-reflex-0.8b-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use deepbot-studio/deepbot-reflex-0.8b with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use deepbot-studio/deepbot-reflex-0.8b with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf deepbot-studio/deepbot-reflex-0.8b:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "deepbot-studio/deepbot-reflex-0.8b:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

DeepBot Reflex 0.8B (r3)

DeepBot's action model: it works a Windows desktop one step at a time - look at the screen, decide the next action, do it - and it is the model that DeepBot's naps keep training.

The base is Qwen3.5-0.8B (Apache-2.0, © Alibaba Qwen). What is published here is what DeepBot trained on top of it: a LoRA (rank 16, alpha 32) trained for two epochs on 14,440 examples of desktop work, merged into the base.

Why this round

Four rounds were trained and each was measured on the same sealed test of 150 live cases, not on the training loss:

round	live test	system	apps	files	flows	refusals	ask	eval loss
r1	107/150	20/24	26/34	25/42	7/20	19/20	10/10	0.0880
r2	114/150	20/24	26/34	31/42	9/20	20/20	8/10	0.0595
r3	114/150	23/24	28/34	27/42	8/20	20/20	8/10	0.0482
r4	111/150	23/24	26/34	27/42	7/20	19/20	9/10	0.0392

r4 has the best loss and the worst live result of the last three rounds, so r3 is what ships. The ceiling on this data is about 76%.

Files

deepbot-reflex-0.8B-r3.Q4_K_M.gguf - Q4_K_M, 516 MiB, run by llama.cpp
model.safetensors with config and tokenizer files - the merged model, for continuing training
adapter_model.safetensors - the LoRA on its own
train_meta.json - what the training run recorded

Run it

llama-server -m deepbot-reflex-0.8B-r3.Q4_K_M.gguf -c 16384 -ngl 99

DeepBot starts it itself and sends the same JSON lines it sends to any Reflex model.

How it keeps learning

A DeepBot takes this as its starting point and naps between tasks: a few small steps at a time, three times as many repeats of older lessons as new ones, at lower process priority, and a task starting stops a nap after the current step. After four naps the student that napped and the one that works sit the same exam - basic skills, the person's own tasks that never enter training, and what was napped. A version that is better and worse nowhere becomes the main one; if it is worse anywhere the nap is dropped.

Licence

Apache-2.0, the licence of the base model. The base model and its name belong to its authors; this repository publishes what DeepBot trained on top of it.

Downloads last month: 88

Safetensors

Model size

0.9B params

Tensor type

F32

BF16

Model tree for deepbot-studio/deepbot-reflex-0.8b

Base model

Qwen/Qwen3.5-0.8B-Base

Finetuned

Qwen/Qwen3.5-0.8B

Finetuned

unsloth/Qwen3.5-0.8B

Adapter

(54)

this model

Adapters

1 model