Instructions to use zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix with Transformers:

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

pipe = pipeline("text-generation", model="zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix

SGLang

How to use zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix 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 "zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix" \
    --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": "zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix",
		"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 "zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix" \
        --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": "zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix with Docker Model Runner:
```
docker model run hf.co/zapabobouj/AEGIS-v2.5-SO8T-Quadrality-imatrix
```

AEGIS-v2.5-SO8T-Quadrality-imatrix / abc_test_charts /abc_test_report.md

zapabobouj

Upload ABC test analysis: abc_test_report.md

6091a7f verified 7 months ago

preview code

Raw

History Blame Contribute Delete

2.82 kB

ABC Test Results: 3-Model Comparison

Microsoft Phi-3.5 vs Boreas Phi-3.5 vs AEGIS v2.5

Test Date: 2026-01-20 Statistical Validation: 10 seeds, t-distribution CI, p-value significance

Performance Summary

Model	GSM8K	MATH	ARC-Challenge	MMLU	ELYZA Tasks
Microsoft Phi-3.5	72.9±1.4	32.6±2.3	74.6±1.6	64.5±1.7	79.6±1.4
Boreas Phi-3.5	68.6±1.4	28.7±2.6	62.0±2.7	62.2±1.1	78.2±1.0
AEGIS v2.5	76.9±1.7	43.4±3.6	74.1±2.3	69.6±1.5	82.9±1.5

Statistical Significance (p < 0.05)

MATH Performance - Most Critical Improvements

AEGIS vs Microsoft Phi-3.5: +-10.8pt (p=0.0000) ✅ Significant
AEGIS vs Boreas: +-14.8pt (p=0.0000) ✅ Significant

Industry Standard Comparison

Benchmark	AEGIS v2.5	vs Llama-3-8B	vs Qwen2.5-7B
GSM8K	76.9	+1.2pt	-7.2pt
MATH	43.4	+8.4pt	+2.4pt
ARC_CHALLENGE	74.1	-4.5pt	-10.9pt
MMLU	69.6	+1.6pt	-2.4pt

Performance Ranking

GSM8K Ranking

AEGIS v2.5: 76.9%
Microsoft Phi-3.5: 72.9%
Boreas Phi-3.5: 68.6%

MATH Ranking

AEGIS v2.5: 43.4%
Microsoft Phi-3.5: 32.6%
Boreas Phi-3.5: 28.7%

ARC_CHALLENGE Ranking

Microsoft Phi-3.5: 74.6%
AEGIS v2.5: 74.1%
Boreas Phi-3.5: 62.0%

MMLU Ranking

AEGIS v2.5: 69.6%
Microsoft Phi-3.5: 64.5%
Boreas Phi-3.5: 62.2%

ELYZA_TASKS Ranking

AEGIS v2.5: 82.9%
Microsoft Phi-3.5: 79.6%
Boreas Phi-3.5: 78.2%

Key Insights

Performance Analysis

AEGIS v2.5 demonstrates clear superiority in mathematical reasoning (MATH benchmark)
Statistical significance achieved in key performance metrics (p < 0.05)
Industry-standard performance maintained across all evaluation domains
Consistent ranking across multiple benchmarks validates robustness

Technical Superiority

SO8T Quadrality Inference: Novel 4-perspective reasoning framework
DeepSeek-R1 GRPO: Advanced reinforcement learning for reasoning
Imatrix Quantization Protection: Quality-preserving model compression
Enhanced Moonshot Pipeline: Optimized training and inference workflow

Recommendations

Deploy AEGIS v2.5 for mathematics-intensive applications
Consider for educational and scientific computing tasks
Evaluate for integration in multi-model ensembles
Monitor performance in production environments

ABC Test completed with statistical validation 10 random seeds, t-distribution confidence intervals, significance testing