Alpie-Core: 4-bit Quantized Reasoning Model
[Space reserved for blog paper, technical report links, and company logo]
1. Introduction
Alpie-Core is one of the world's first fine-tuned 4-bit reasoning models, proving that aggressive quantization can surpass full-precision baselines in reasoning, mathematics, and coding. By combining cutting-edge quantization-aware training with synthetic STEM-rich datasets, Alpie-Core achieves frontier-level reasoning while being practical for real-world deployment at scale.
2. Model Summary
- Base Architecture: DeepSeek-R1-Distill-Qwen-32B
- Parameters: 32 billion (quantized to 4-bit)
- Training Method: Supervised Fine-Tuning (SFT) using LoRA/QLoRA techniques
- Quantization: 4-bit NF4 with double quantization
- Context Length: 65,536 tokens
- Max Output Length: 16,384 tokens
- License: Apache 2.0
- Memory Footprint: ~8GB (75% reduction from full-precision)
3. Model Features
- Supports Streaming – Real-time token-level responses
- OpenAI-Compatible API – Seamless integration with OpenAI client libraries
- 65K Context Length – Handles very large inputs and conversations
- 16,384 Max Output Length – Enables extremely long generations
- 4-Bit Quantization – Memory-efficient and optimized for deployment
- High Throughput Inference – Powered by vLLM for efficient large-scale serving
- Low Latency Inference – Fast response times optimized for production
- Customizable Safety & Moderation Filters – Built-in guardrails for safer outputs
- Supports Function Calling / Tool Use – Enables structured outputs and external API integration
4. Key Highlights
- Frontier Performance in 4-bit: 81.28% MMLU, 92.75% GSM8K, 57.8% SWE-Bench Verified
- Global Ranking: 3rd place on Humanity's Last Exam leaderboard
- Cost Advantage: 70-88% lower inference cost vs GPT-4/Claude/DeepSeek
- Environmental Impact: 64% lower carbon footprint per inference
- STEM + Coding Excellence: Outperforms full-precision peers in mathematics and programming
- Enhanced Content Access: Provides factual responses to geopolitically sensitive topics
5. Benchmark Results
| Benchmark | Alpie-Core (32B-4bit) | DeepSeek-V2 (236B) | Qwen2.5 72B | Llama 3.1 405B | Llama 3.1 70B | Gemma-3 27B-PT | Mistral-Small-24B-Base-2501 |
|---|---|---|---|---|---|---|---|
| MMLU (5-shot) | 81.28% | 78.4% | 85.0% | 84.4% | 79.3% | 78.6% | 80.73% |
| GSM8K (8-shot) | 92.75% | 81.6% | 88.3% | 83.5% | nan | 82.2% | 80.73% |
| BBH (3-shot) | 85.12% | 78.8% | 79.8% | 82.9% | 81.6% | 77.7% | nan |
| MMLU-Pro (5-shot) | 64.78% | 51.4% | 58.3% | 52.8% | 53.8% | 52.2% | 54.37% |
| MBPP (pass@1) | 75.20% | 65.0% | 72.6% | 68.4% | nan | 65.6% | 69.64% |
| HumanEval (pass@1) | 57.23% | 43.3% | 53.0% | 54.9% | nan | 48.8% | nan |
SWE-Bench Verified Performance
| Rank | Model | Accuracy (%) | Performance vs Alpie |
|---|---|---|---|
| 1 | Alpie Core | 57.8 | Alpie |
| 2 | Qwen3-Coder-30B-A3B-Instruct | 51.6 | Below Alpie |
| 3 | o1 | 48.9 | Below Alpie |
| 4 | o3-mini (high) | 49.3 | Below Alpie |
| 5 | Claude 3.5 Sonnet | 49.0 | Below Alpie |
| 6 | DeepSeek R1 | 49.2 | Below Alpie |
| 7 | Devstral | 46.8 | Below Alpie |
Humanity's Last Exam Leaderboard Performance
| Rank | Model | Accuracy (%) | Performance vs Alpie |
|---|---|---|---|
| 1 | GPT 4.5 Preview | 5.8 | Above Alpie |
| 2 | Claude Sonnet 4 | 5.42 | Above Alpie |
| 3 | Alpie Core 32B (4-bit) | 5.41 | Alpie |
| 4 | Llama 4 Maverik | 5.34 | Below Alpie |
| 5 | GPT 4.1 | 4.97 | Below Alpie |
| 6 | Kimi K2 Instruct | 4.68 | Below Alpie |
| 7 | DeepSeek V3 | 4.55 | Below Alpie |
| 8 | Gemini 1.5 Pro 002 | 4.55 | Below Alpie |
Additional Benchmarks
| Benchmark | Alpie-Core (32B-4bit) | Category |
|---|---|---|
| AIME | 47.34% | Advanced Mathematics |
| GPQA (Diamond) | 40.91% | Graduate-level QA |
| TruthfulQA (MC2) | 60.05% | Truthfulness |
| HellaSwag | 84.66% | Commonsense |
| PIQA | 83.24% | Physical Reasoning |
| ARC Challenge | 67.58% | Science QA |
| CommonSenseQA | 87.06% | Commonsense |
| AGIEval | 64.98% | General Intelligence |
| Winogrande | 79.53% | Commonsense Reasoning |
6. Training Details
- Hardware: 8× NVIDIA A100-80GB GPUs
- Training Duration: 408 hours
- Fine-tuning Method: LoRA/QLoRA with the following configuration:
- LoRA Alpha: 8
- LoRA Dropout: 0.05
- LoRA Rank: 8
- Quantization: 4-bit NF4 + Double Quantization + FP16 compute
- Dataset Domains: Mathematics, coding, reasoning, science, general knowledge, competitive exams, Indian context + law, multilingual (Hindi and Hinglish)
- Synthetic Data Advantage: +15-20% performance boost in STEM & coding domains
7. Environmental Impact
Carbon Footprint: 298-835 kg CO₂e (training)
8. Use Cases
Scientific Research Excellence
- 98% performance on SciQ benchmark
- Advanced physics, chemistry, and mathematical sciences
- Literature review automation and hypothesis generation
- Experimental design optimization
Advanced Coding and Software Engineering
- 57.8% SWE-Bench Verified score (8% above nearest competitor)
- Automated bug detection and GitHub issue resolution
- Competitive programming and algorithm design
- Enterprise software development and architecture design
Indian Cultural and Religious Expertise
- Comprehensive understanding of Hindu philosophy, Buddhist traditions
- Regional diversity and cultural knowledge across Indian states
- Legal and constitutional framework understanding
- Educational support for Indian competitive exams (JEE, NEET, UPSC, SSC)
9. Safety and Limitations
Enhanced Content Access
Unlike the base DeepSeek model, Alpie-Core provides factual, balanced responses to geopolitically sensitive questions, offering global accessibility and factual accuracy on topics like Taiwan's status, Arunachal Pradesh sovereignty, and other sensitive geopolitical issues.
Current Limitations
- Multilingual reasoning in Hindi/Hinglish shows room for improvement
- Fixed knowledge cutoff without real-time information retrieval
- Occasional struggles with complex multi-hop mathematical reasoning
- Potential hallucinations in factual question-answering
Mitigations
- Safety classifiers and output filtering systems
- Model-assisted safety pipeline using RLHF
- Comprehensive adversarial testing by domain experts
10. How to Use
Non-Streaming Inference
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel, PeftConfig
import torch
# Load LoRA adapter configuration to find the base model
peft_model_id = "169Pi/Alpie-core"
config = PeftConfig.from_pretrained(peft_model_id)
# Load the base model
base_model = AutoModelForCausalLM.from_pretrained(
config.base_model_name_or_path,
torch_dtype=torch.float16,
device_map="auto"
)
# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
# Load LoRA weights
model = PeftModel.from_pretrained(base_model, peft_model_id)
# Ensure evaluation mode
model.eval()
# Sample inference
prompt = "Solve the Riemann Hypothesis and provide a final answer?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
with torch.no_grad():
outputs = model.generate(**inputs, max_new_tokens=1000)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print("Response:\n", response)
Streaming Inference
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
from peft import PeftModel, PeftConfig
import torch
# Load LoRA adapter configuration to find the base model
peft_model_id = "169Pi/Alpie-core"
config = PeftConfig.from_pretrained(peft_model_id)
# Load the base model
base_model = AutoModelForCausalLM.from_pretrained(
config.base_model_name_or_path,
torch_dtype=torch.float16,
device_map="auto"
)
# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
# Load LoRA weights
model = PeftModel.from_pretrained(base_model, peft_model_id)
# Ensure evaluation mode
model.eval()
# Initialize streamer
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
# Sample streaming inference
prompt = "Solve the Riemann Hypothesis and provide a final answer?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
print("Streaming Response:")
with torch.no_grad():
outputs = model.generate(
**inputs,
max_new_tokens=1000,
streamer=streamer,
do_sample=True,
temperature=0.7,
top_p=0.9
)
Deployment Options
- Transformers: Python, PyTorch integration
- vLLM: High-throughput inference
- LMDeploy/Ollama/TensorRT-LLM: Production deployments
11. Citation
@misc{alpie2025core,
title = {Alpie-Core: A 4-bit Quantized Reasoning Model Surpassing Full-Precision Benchmarks},
author = {Alpie AI},
year = {2025},
url = {https://huggingface.co/alpie/Alpie-Core-4bit}
}
12. License
Apache 2.0 – Free for research and commercial use
For technical details, training methodology, and comprehensive evaluation results, please refer to our technical report.