Skywork-OR1-7B GGUF
GGUF quantizations of Skywork/Skywork-OR1-7B.
Description
Skywork-OR1-7B (Open Reasoner 1) is a general-purpose math and code reasoning model trained with large-scale rule-based reinforcement learning using a customized GRPO algorithm. It is based on DeepSeek-R1-Distill-Qwen-7B and trained on 110K math problems and 14K coding questions with model-aware difficulty estimation, offline/online filtering, rejection sampling, multi-stage training pipeline, and adaptive entropy control.
Quantization Files
| File | Bits | Size | Use Case |
|---|---|---|---|
| model_f16.gguf | 16 | ~15.2 GB | Maximum quality, reference |
| model_q8_0.gguf | 8 | ~8.5 GB | Best quality, near-lossless |
| model_q6_k.gguf | 6 | ~6.6 GB | High quality |
| model_q5_k_m.gguf | 5 | ~5.7 GB | Balanced quality |
| model_q5_k_s.gguf | 5 | ~5.5 GB | Balanced quality, smaller |
| model_q4_k_m.gguf | 4 | ~4.9 GB | Good quality, recommended |
| model_q4_k_s.gguf | 4 | ~4.7 GB | Good quality, smaller |
| model_q3_k_l.gguf | 3 | ~4.0 GB | Low quality, small |
| model_q3_k_m.gguf | 3 | ~3.7 GB | Low quality, smaller |
| model_q3_k_s.gguf | 3 | ~3.5 GB | Very low quality |
| model_q2_k.gguf | 2 | ~3.0 GB | Lowest quality, minimum size |
VRAM Requirements
| Quantization | VRAM |
|---|---|
| F16 | ~16 GB |
| Q8_0 | ~9 GB |
| Q6_K | ~7 GB |
| Q5_K_M | ~6 GB |
| Q4_K_M | ~5 GB |
| Q3_K_M | ~4 GB |
| Q2_K | ~3.5 GB |
Usage
llama.cpp
./llama-cli -m model_q4_k_m.gguf -p "<your prompt>" -n 512
llama-cpp-python
from llama_cpp import Llama
llm = Llama(model_path="model_q4_k_m.gguf", n_ctx=32768)
output = llm("<your prompt>", max_tokens=512)
print(output["choices"][0]["text"])
LM Studio
Load any .gguf file directly in LM Studio via Load Model.
Ollama
ollama run hf.co/tinyopsec/Skywork-OR1-7B-GGUF:Q4_K_M
Original Model
- Skywork/Skywork-OR1-7B
- Training data: Skywork/Skywork-OR1-RL-Data
- License: Apache 2.0