Instructions to use pplgltch/QUEST-35B-RL-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use pplgltch/QUEST-35B-RL-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 pplgltch/QUEST-35B-RL-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf pplgltch/QUEST-35B-RL-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 pplgltch/QUEST-35B-RL-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf pplgltch/QUEST-35B-RL-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 pplgltch/QUEST-35B-RL-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf pplgltch/QUEST-35B-RL-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 pplgltch/QUEST-35B-RL-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf pplgltch/QUEST-35B-RL-GGUF:Q4_K_M

Use Docker

docker model run hf.co/pplgltch/QUEST-35B-RL-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use pplgltch/QUEST-35B-RL-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "pplgltch/QUEST-35B-RL-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": "pplgltch/QUEST-35B-RL-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/pplgltch/QUEST-35B-RL-GGUF:Q4_K_M

Ollama
How to use pplgltch/QUEST-35B-RL-GGUF with Ollama:
```
ollama run hf.co/pplgltch/QUEST-35B-RL-GGUF:Q4_K_M
```

Unsloth Studio

How to use pplgltch/QUEST-35B-RL-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 pplgltch/QUEST-35B-RL-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 pplgltch/QUEST-35B-RL-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for pplgltch/QUEST-35B-RL-GGUF to start chatting

How to use pplgltch/QUEST-35B-RL-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf pplgltch/QUEST-35B-RL-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": "pplgltch/QUEST-35B-RL-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use pplgltch/QUEST-35B-RL-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 pplgltch/QUEST-35B-RL-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 pplgltch/QUEST-35B-RL-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use pplgltch/QUEST-35B-RL-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf pplgltch/QUEST-35B-RL-GGUF: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 "pplgltch/QUEST-35B-RL-GGUF: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"

Docker Model Runner
How to use pplgltch/QUEST-35B-RL-GGUF with Docker Model Runner:
```
docker model run hf.co/pplgltch/QUEST-35B-RL-GGUF:Q4_K_M
```

Lemonade

How to use pplgltch/QUEST-35B-RL-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull pplgltch/QUEST-35B-RL-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.QUEST-35B-RL-GGUF-Q4_K_M

List all available models

lemonade list

QUEST-35B-RL — GGUF

GGUF quantizations of osunlp/QUEST-35B-RL, the 35B RL-trained deep-research agent from the paper QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks (built on Qwen3.5-35B-A3B, MoE with ~3B active parameters).

Files

file	size	contents
`QUEST-35B-RL-Q8_0.gguf`	35 GB	Q8_0 — near-lossless
`QUEST-35B-RL-Q6_K.gguf`	~27 GB	Q6_K
`QUEST-35B-RL-Q5_K_M.gguf`	~23 GB	Q5_K_M
`QUEST-35B-RL-Q4_K_M.gguf`	~19 GB	Q4_K_M
`QUEST-35B-RL-mmproj-F16.gguf`	858 MB	vision projector (F16), optional — only needed for image input

K-quants were made from a BF16 intermediate GGUF (not requantized from Q8_0), no imatrix.

Usage

Requires a llama.cpp build with qwen35moe support (mid-2026 or newer).

llama-server -m QUEST-35B-RL-Q8_0.gguf \
  --mmproj QUEST-35B-RL-mmproj-F16.gguf \
  -ngl 99 --ctx-size 65536

The model is a deep-research agent: it is trained for long-horizon, tool-using search trajectories (fact-seeking, citation grounding, report synthesis) and performs best behind an agent harness with search/browse tools, rather than as a plain chatbot. Thinking mode (<think>) is native. The architecture supports up to 262k context. The model card of the base repo specifies no sampling parameters; Qwen3.5 reasoning defaults (temperature ≈ 0.6, top_p ≈ 0.95) are a sensible starting point.

Conversion notes

Converted with convert_hf_to_gguf.py (llama.cpp commit e3546c7), --outtype q8_0, chat template embedded from the source repo. Two source-repo quirks were handled during conversion:

--no-mtp was required. The checkpoint's config.json declares mtp_num_hidden_layers: 1 (multi-token prediction, inherited from the Qwen3.5 base), but the RL finetune contains no mtp.* tensors. Converting with defaults produces a GGUF that fails to load with missing tensor 'blk.40.attn_norm.weight'.
The repo was saved with transformers v5 and declares the v5-only tokenizer class TokenizersBackend; tokenizer_config.json was patched to the equivalent PreTrainedTokenizerFast for conversion.

Verified after conversion: loads and generates coherently (temp 0 smoke test), 74 t/s generation fully offloaded on 3× AMD R9700 (ROCm).

License

Apache 2.0, same as the source model. All credit for the model to the OSU NLP group; this repo only provides a format conversion.

Downloads last month: 58

GGUF

Model size

35B params

Architecture

qwen35moe

Hardware compatibility

4-bit

5-bit

6-bit

8-bit

Model tree for pplgltch/QUEST-35B-RL-GGUF

Base model

osunlp/QUEST-35B-RL

Quantized

(4)

this model

Paper for pplgltch/QUEST-35B-RL-GGUF

QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks

Paper • 2605.24218 • Published May 22 • 46