Instructions to use ModernoTech/Nova-Core with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use ModernoTech/Nova-Core 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 ModernoTech/Nova-Core:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ModernoTech/Nova-Core:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ModernoTech/Nova-Core:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ModernoTech/Nova-Core:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf ModernoTech/Nova-Core:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf ModernoTech/Nova-Core:Q4_K_M

Use Docker

docker model run hf.co/ModernoTech/Nova-Core:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use ModernoTech/Nova-Core with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/ModernoTech/Nova-Core:Q4_K_M

Ollama
How to use ModernoTech/Nova-Core with Ollama:
```
ollama run hf.co/ModernoTech/Nova-Core:Q4_K_M
```
Unsloth Desktop

How to use ModernoTech/Nova-Core with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ModernoTech/Nova-Core:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "ModernoTech/Nova-Core:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use ModernoTech/Nova-Core with Docker Model Runner:
```
docker model run hf.co/ModernoTech/Nova-Core:Q4_K_M
```

Lemonade

How to use ModernoTech/Nova-Core with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ModernoTech/Nova-Core:Q4_K_M

Run and chat with the model

lemonade run user.Nova-Core-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use ModernoTech/Nova-Core with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ModernoTech/Nova-Core: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 ModernoTech/Nova-Core:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ModernoTech/Nova-Core with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ModernoTech/Nova-Core: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 "ModernoTech/Nova-Core: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"

ModernoTech commited on 5 days ago

Commit

ef00749

verified ·

1 Parent(s): 4752068

Add quantitative performance tables, status labels and verified chart references

Browse files

Files changed (1) hide show

README.md +65 -2

README.md CHANGED Viewed

@@ -41,6 +41,69 @@ Modelo denso conversacional basado en Gemma 4, documentado bajo la identidad Nov
   - Prompt: 49.5 t/s
   - Generación: 34.5 t/s
 ## Limitaciones
 - El consumo real depende de `-ngl` y del contexto usado.
@@ -57,6 +120,6 @@ MODERNO TECH · Nova AI · Intelligence. Engineering. Innovation.
 ## Demostración visual de inferencia
-![Animación de inferencia](images/{short.lower()}-inference-process.gif)
 ![Animación de inferencia](images/nova-core-inference-process.gif)

   - Prompt: 49.5 t/s
   - Generación: 34.5 t/s
+## Rendimiento cuantitativo
+> Todos los valores de throughput son **ESTIMADOS** (techo de ancho de banda) o **NO VALIDADOS**. Ver `performance/performance.json` y `performance/simulation-results.csv`.
+Única medición documentada: MacBook Air M4 16 GB, contexto 512, carga 3.16 s, prompt 49.5 tok/s, decode 34.5 tok/s (flags -ngl 0, --ctx-size 512, --temp 0.7).
+- Artefacto: binario verificado de 3,009,448,896 bytes
+- Contextos analizados: 128K, 4K, 64K, 8K.
+- MNE: proyección SSD a 2.0 GiB/s, cache 90 %–99.9 %, KV a FP16, payload Q4 activo.
+### Memoria y throughput por hardware (contexto 4K, estimado)
+| Hardware | RAM disponible (GB) | Ancho banda (GiB/s) | Límite RAM tok/s | MNE tok/s a 99% hit |
+|---|---:|---:|---:|---:|
+| MacBook Air 2014 | 4 | 25.6 | 10.1 | 79.1 |
+| Lenovo Yoga 720 | 8 | 34.1 | 13.5 | 79.1 |
+| MacBook Neo A18 Pro | 8 | 60.0 | 23.7 | 79.1 |
+| MacBook Air M4 | 16 | 120.0 | 47.4 | 79.1 |
+### Escenarios MNE de cache
+| Cache hit | Miss leído desde SSD | Techo tok/s (modelo/representante) |
+|---|---:|---:|
+| 90.0 % | 10.0 % | 7.9 |
+| 95.0 % | 5.0 % | 15.8 |
+| 99.0 % | 1.0 % | 79.1 |
+| 99.5 % | 0.5 % | 158.1 |
+| 99.9 % | 0.1 % | 790.5 |
+### Contexto nativo y KV
+- Máximo posición documentado: 131,072 tokens.
+- KV se estima por `contexto × layers × kv_heads × head_dim × 2 (K+V) × 2 bytes`.
+- Para contextos superiores al nativo se marca `NO VALIDADO` en los CSV/JSON.
+### Gráficos
+![Tokens por segundo (límite teórico RAM)](performance/tokens-per-second.png)
+![RAM vs MNE (proyección)](performance/ram-vs-mne.png)
+![Contexto vs memoria KV (estimado)](performance/context-performance.png)
+![Distribución de memoria (estimado)](performance/memory-breakdown.png)
+![Tamaño vs RAM disponible](performance/model-size-vs-ram.png)
+![Sensibilidad del cache MNE](performance/cache-hit-analysis.png)
+![Prefill vs decode](performance/prefill-vs-decode.png)
+### Archivos publicados
+- `performance/performance.json`
+- `performance/simulation-results.csv`
+- `performance/methodology.md`
+- `performance/hardware-reference.md`
+- `performance/generate_charts.py`
+- `performance/generate_animation.py`
+- Ver `performance/` para los 7 PNG de esta fase.
 ## Limitaciones
 - El consumo real depende de `-ngl` y del contexto usado.
 ## Demostración visual de inferencia
 ![Animación de inferencia](images/nova-core-inference-process.gif)
+![Animación de inferencia](images/nova-core-inference-process.gif)