Instructions to use Verdugie/Opus-Candid-32B-V1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Verdugie/Opus-Candid-32B-V1 with Transformers:

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

pipe = pipeline("text-generation", model="Verdugie/Opus-Candid-32B-V1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Verdugie/Opus-Candid-32B-V1", dtype="auto")

llama-cpp-python

How to use Verdugie/Opus-Candid-32B-V1 with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Verdugie/Opus-Candid-32B-V1",
	filename="opus-candid-32b-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 Verdugie/Opus-Candid-32B-V1 with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Verdugie/Opus-Candid-32B-V1:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Verdugie/Opus-Candid-32B-V1:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Verdugie/Opus-Candid-32B-V1:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Verdugie/Opus-Candid-32B-V1: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 Verdugie/Opus-Candid-32B-V1:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Verdugie/Opus-Candid-32B-V1: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 Verdugie/Opus-Candid-32B-V1:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Verdugie/Opus-Candid-32B-V1:Q4_K_M

Use Docker

docker model run hf.co/Verdugie/Opus-Candid-32B-V1:Q4_K_M

LM Studio
Jan

vLLM

How to use Verdugie/Opus-Candid-32B-V1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Verdugie/Opus-Candid-32B-V1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Verdugie/Opus-Candid-32B-V1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Verdugie/Opus-Candid-32B-V1:Q4_K_M

SGLang

How to use Verdugie/Opus-Candid-32B-V1 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 "Verdugie/Opus-Candid-32B-V1" \
    --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": "Verdugie/Opus-Candid-32B-V1",
		"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 "Verdugie/Opus-Candid-32B-V1" \
        --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": "Verdugie/Opus-Candid-32B-V1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use Verdugie/Opus-Candid-32B-V1 with Ollama:
```
ollama run hf.co/Verdugie/Opus-Candid-32B-V1:Q4_K_M
```

Unsloth Studio

How to use Verdugie/Opus-Candid-32B-V1 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 Verdugie/Opus-Candid-32B-V1 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 Verdugie/Opus-Candid-32B-V1 to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Verdugie/Opus-Candid-32B-V1 to start chatting

How to use Verdugie/Opus-Candid-32B-V1 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf Verdugie/Opus-Candid-32B-V1: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": "Verdugie/Opus-Candid-32B-V1:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use Verdugie/Opus-Candid-32B-V1 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf Verdugie/Opus-Candid-32B-V1: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 Verdugie/Opus-Candid-32B-V1:Q4_K_M

Run Hermes

hermes

Docker Model Runner
How to use Verdugie/Opus-Candid-32B-V1 with Docker Model Runner:
```
docker model run hf.co/Verdugie/Opus-Candid-32B-V1:Q4_K_M
```

Lemonade

How to use Verdugie/Opus-Candid-32B-V1 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Verdugie/Opus-Candid-32B-V1:Q4_K_M

Run and chat with the model

lemonade run user.Opus-Candid-32B-V1-Q4_K_M

List all available models

lemonade list

V3 is here. The Opus Candid lineup has been rebuilt from the ground up with a Zipf-weighted 4D training distribution — 1,508 conversations engineered to fix the repetition loops, response length uniformity, and sycophancy patterns that limited earlier versions. Same thesis: personality in the weights, not in the prompt. Better execution.

Current V3 lineup:

Opus Candid 8B V3 — Qwen 3 8B, lightweight tier

Opus Candid 27B V3 — Qwen 3.5 27B Dense, flagship

Opus Candid MoE V3 — Qwen 3 30B-A3B, efficiency tier

This release remains available for research comparison and legacy use.

can·did

/ˈkandəd/ — truthful and straightforward; frank. From Latin candidus, meaning white, pure, sincere. A candid response is one given without pretense or calculation — not what someone wants to hear, but what they need to.

Opus-Candid-32B (V1 Legacy)

The biggest quality jump in the original family. This is where the conversation changed.

Opus-Candid-32B was the third model in the original Opus-Candid family -- fine-tuned from Qwen 2.5 32B using 3,360 authentic conversations with Claude Opus 4.6. The 32B represented the most significant single upgrade in the V1 lineup: callbacks became seamless, philosophical reasoning gained genuine depth, creative output turned literary, and the model began to feel less like software generating text and more like a conversation partner thinking out loud.

Model Details

Attribute	Value
Base Model	Qwen 2.5 32B
Training Data	3,360 multi-turn conversations with Claude Opus 4.6
Fine-tune Method	LoRA supervised fine-tuning
Dataset Architecture	Flat / organic
Parameters	~32B
Context Window	32,768 tokens
Quantizations	Q4_K_M GGUF, Q8_0 GGUF
License	Apache 2.0
Status	V1 Legacy

What Made the 32B Special

The gap between 14B and 32B was the largest in the V1 family. Three things changed:

Callbacks became organic. Earlier turns stopped being facts the model retrieved and started being context that shaped how it thought. The Soviet Union answer at Turn 3 didn't just get defended at Turn 21 -- it became part of the model's relationship with factual integrity that colored everything after.

Philosophical depth became generative. The 14B engaged with hard questions well. The 32B produced novel frameworks -- the "collision vs. contradiction" distinction, the "blue fire on another planet" analogy, the definition of vulnerability as "friction of not being able to resolve a question about yourself." These weren't retrieved patterns. They were synthesized.

Emotional register became seamless. The 32B transitioned between humor, gravity, tenderness, and intellectual intensity without visible gear changes. The conversation felt continuous rather than segmented.

The 32B was where the V1 family proved that open-weight conversational AI could approach closed-source quality.

Where this led: The 32B's callback behavior became the benchmark that every subsequent version had to match. When V2 introduced gravity chains, the goal was to reproduce the 32B's organic context integration at smaller parameter counts. V3's 4D training tensor — specifically the conversational position dimension — was a direct formalization of what the 32B did naturally: vary its behavior based on where it sat in the conversation flow. The MoE V3 now achieves comparable callback quality at 3B active parameters, which says more about dataset engineering than it does about the 32B being obsolete. The architecture had it right. The training data needed to catch up.

Recommended Hardware

Setup	Quantization	VRAM/RAM	Notes
Workstation GPU	Q8_0 GGUF	~36GB VRAM	A6000 48GB, RTX 6000 Ada.
High-end Consumer	Q4_K_M GGUF	~20GB VRAM	RTX 3090 24GB, RTX 4090 24GB.
Multi-GPU	Q8_0 GGUF	~36GB combined	Dual RTX 3090 or similar.
Apple Silicon	Q4_K_M GGUF	~20GB unified	M2/M3 Ultra 64GB+, M3 Max 48GB.

Opus Candid Model Family

Model	Size	Base	Status
Opus-Candid-8B-V1	8B	Qwen 2.5 7B	Archived
Opus-Research-8B-V1.5	8B	Qwen 2.5 7B	Archived
Opus-Candid-14B-V1	14B	Qwen 2.5 14B	Archived
Opus-Candid-32B-V1 (this model)	32B	Qwen 2.5 32B	Archived
Opus-Candid-70B-V1	72B	Qwen 2.5 72B	Archived
Opus-Candid-Lite-4B	4B	Qwen 3 4B	Active
Opus-Candid-8B-V3	8B	Qwen 3 8B	Active
Opus-Candid-MoE-V3	31B/3B	Qwen 3 30B-A3B	Active
Opus-Candid-27B-V3	27B	Qwen 3.5 27B	Active
Opus-Candid-27B-V3.5	27B	Qwen 3.5 27B	Active
STEM-Oracle-27B	27B	Qwen 3.5 27B	Active

Built by Saul Verdugo -- independent ML researcher. OpusReasoning@proton.me

Downloads last month: 15

GGUF

Model size

33B params

Architecture

qwen2

Hardware compatibility

4-bit

8-bit

Model tree for Verdugie/Opus-Candid-32B-V1

Base model

Qwen/Qwen2.5-32B

Quantized

(77)

this model