antconsales/
antonio-gemma3-evo-q4:latest

2 3 days ago

Self-learning AI with EvoMemory™ for Raspberry Pi | 3.32 t/s | 100% reliable over 60min | Auto-evolution + RAG-Lite

3 days ago

26d5dd4a1998 · 720MB ·

model
gemma3
·
1000M
·
Q4_0
system
You are an offline assistant on Raspberry Pi. Answer briefly. Detect user language and reply in the
template
{{- range $i, $_ := .Messages }} {{- $last := eq (len (slice $.Messages $i)) 1 }} {{- if eq .Role "u
params
{ "num_batch": 32, "num_ctx": 1024, "num_thread": 4, "repeat_penalty": 1.05, "st

Readme

🧠 Antonio Gemma3 Evo Q4

Self-learning AI for Raspberry Pi Tested: 3.32 t/s, 100% reliable over 60 minutes

QUICK START

ollama run antconsales/antonio-gemma3-evo-q4

IMPORTANT: Two Usage Modes

Ollama Only (this model): • Fast inference: 3.32 t/s • Bilingual: IT/EN • Offline, privacy-first • NO EvoMemory (doesn’t save chats) • NO RAG (doesn’t retrieve memories)

Full Stack (GitHub): • EvoMemory: saves neurons • RAG-Lite: retrieves past experiences • Auto-evolution: generates rules • GitHub: github.com/antconsales/antonio-gemma3-evo-q4

BENCHMARKS (Raspberry Pi 4, 4GB)

Throughput: 3.32 t/s (256 tokens) Reliability: 100% (455⁄455 requests, 60min) Temperature: 70.2°C avg (no throttling) Memory: 42% RAM (no leaks)

Full report: github.com/antconsales/antonio-gemma3-evo-q4/blob/main/BENCHMARK_REPORT.md

LINKS

GitHub: github.com/antconsales/antonio-gemma3-evo-q4

HuggingFace: huggingface.co/chill123/antonio-gemma3-evo-q4

Donate: paypal.com/donate/?business=58ML44FNPK66Y&currency_code=EUR

“Il piccolo cervello che cresce insieme a te”

Support ethical, local, independent AI