Instructions to use deepreinforce-ai/Ornith-1.0-35B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with Transformers:

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

pipe = pipeline("text-generation", model="deepreinforce-ai/Ornith-1.0-35B-GGUF")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("deepreinforce-ai/Ornith-1.0-35B-GGUF", dtype="auto")

llama-cpp-python

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="deepreinforce-ai/Ornith-1.0-35B-GGUF",
	filename="ornith-1.0-35b-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF 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 deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf deepreinforce-ai/Ornith-1.0-35B-GGUF: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 deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf deepreinforce-ai/Ornith-1.0-35B-GGUF: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 deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deepreinforce-ai/Ornith-1.0-35B-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "deepreinforce-ai/Ornith-1.0-35B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M

SGLang

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF 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 "deepreinforce-ai/Ornith-1.0-35B-GGUF" \
    --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": "deepreinforce-ai/Ornith-1.0-35B-GGUF",
		"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 "deepreinforce-ai/Ornith-1.0-35B-GGUF" \
        --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": "deepreinforce-ai/Ornith-1.0-35B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with Ollama:
```
ollama run hf.co/deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M
```

Unsloth Studio

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for deepreinforce-ai/Ornith-1.0-35B-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for deepreinforce-ai/Ornith-1.0-35B-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for deepreinforce-ai/Ornith-1.0-35B-GGUF to start chatting

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf deepreinforce-ai/Ornith-1.0-35B-GGUF: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 deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with Docker Model Runner:
```
docker model run hf.co/deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M
```

Lemonade

How to use deepreinforce-ai/Ornith-1.0-35B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Ornith-1.0-35B-GGUF-Q4_K_M

List all available models

lemonade list

After loading this model, the agent falls into an infinite loop during tool calls, or the response gets cut off within the first three turns.

#13

by AntonioWen - opened 5 days ago

Discussion

AntonioWen

5 days ago

After testing this model with multiple AI agents via llama.cpp, I encountered a few issues:
Besides getting truncated during tool calls in the early turns, it easily falls into infinite loops when executing long tasks.
It occasionally performs well on short tasks.
I hope to see improvements in the future. Based on benchmark evaluations, this is a highly anticipated model, and in my local testing, it has even outperformed MiniMax-M3.

renatomoselli

5 days ago

After testing this model with multiple AI agents via llama.cpp, I encountered a few issues:
Besides getting truncated during tool calls in the early turns, it easily falls into infinite loops when executing long tasks.
It occasionally performs well on short tasks.
I hope to see improvements in the future. Based on benchmark evaluations, this is a highly anticipated model, and in my local testing, it has even outperformed MiniMax-M3.

after struggling with infinite loops on qwen models for a while, I finally found the definitive fix (on windows .bat): --chat-template-kwargs "{"preserve_thinking":true}"

ajeetcoolkarni

4 days ago

I am struggling with same infinite loop, so you say setting --chat-template-kwargs "{"preserve_thinking":true}" solves the issue? Thanks

renatomoselli

4 days ago

yeah, but sadly after a lot of tests the issue happened again, so it just got better. I'm also experimenting with different chat templates and other settings, mmap/no-mmap also seems to have a certain relevance. So it's not a fix, but it got better. Lmk if you manage to find the answer.

QuestionableCube

4 days ago

For me, Ornith is getting stuck much less often than Qwen3.6-35b-a3b. I have been running the Q8_0 quantization all day. It has been making progress all day without any real issues. I am using VS Code 1.124.0 with GitHub Copilot LLM Gateway. It feels like Copilot is better now at keeping the model on the rails - which could just be Ornith being smarter than Qwen. This morning I increased the MaxOutputTokens in LLM Gateway to 65,536 - which also possibly helped.

ajeetcoolkarni

4 days ago

Thanks I'm also trying few things myself I'll update if I find a solution.

QuestionableCube

3 days ago

•

edited 3 days ago

Try the enhanced jinja file for Qwen3.6 to fix issues - see

I also just disabled "Parallel Tool Calling". The model was seeing artifacts when reading long files, e.g. {"$mid":24,"mimeType":"cache_control","data":"..."}. There is a "Parallel Tool Calling" flag in the settings for "GitHub Copilot LLM Gateway" which I am using to bridge between Copilot and Llama.cpp. Disabling that flag, as suggested by Google, seems to have fixed the issue.

Selected llama-server (llama.cpp) arguments:

                --jinja \
                --chat-template-file "${JINJA_FIX_DIR}/qwen3.6-enhanced.jinja" \
                --reasoning-format deepseek \
                --preserve-thinking true \

renatomoselli

3 days ago

--reasoning-format deepseek -> was the final fix for me, I already tested it yesterday all the day and it worked fine, was about to come here and share the news, thankfully you found it out too!!! this issue was super annoying.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment