Complete Mac-first on-device RAG stack — chat LLM + reranker + text/VL embedder, direct from BF16, BatiAI-signed. For BatiFlow.
AI & ML interests
On-device AI, GGUF quantization, Apple Silicon, macOS automation
Recent Activity
Organization Card
BatiAI — Work Less, More Impact
We help individuals and businesses maximize productivity through AI automation — so they can create greater value for the world.
BatiFlow — On-Device AI for Everyone
Free. Unlimited. 100% Private. Just 5MB.
BatiFlow is a macOS-native AI desktop automation app. It connects on-device AI (via Ollama) to your Mac's apps — KakaoTalk, iMessage, Slack, Chrome, Calendar, Notes, and 57+ built-in tools — all controlled by natural language.
Why BatiFlow?
| Free & Unlimited | On-device AI via Ollama — no API costs, no usage limits, no subscriptions |
| 100% Private | All data stays on your Mac. Nothing is sent to the cloud |
| Ultra Lightweight | Native macOS app, only 5MB. No Electron, no bloat |
| 5-Minute Setup | Download, install Ollama, start automating |
| Remote AI via Telegram | Connect on-device AI to Telegram, Discord, Slack bots — full desktop automation from anywhere |
| Multilingual | Korean and English UI + AI responses |
How It Works
AI understands natural language, calls the right tool, and executes the action — messaging, browsing, file management, scheduling, and more. No coding required.
BatiAI Models — Verified GGUF Quantizations
We quantize popular open models directly from official weights using the latest llama.cpp, verify them on real Mac hardware (Mac mini M4 16GB + MacBook Pro M4 Max 128GB), and publish to Ollama and HuggingFace.
Why not third-party quantizations?
We tried. They broke. Third-party GGUF files failed on Ollama 0.20+ due to tensor format mismatches. So we built our own pipeline — from official weights, with imatrix optimization, tested on real consumer hardware.
Available Models
| Model | Ollama | Sizes | Recommended For |
|---|---|---|---|
| Gemma 4 26B MoE | ollama pull batiai/gemma4-26b:iq3 |
IQ3 12GB, Q3 12GB, Q4 16GB | 24GB+ Mac |
Bati CIS — K-Beauty Commerce Intelligence
Our B2B business focuses on K-Beauty global commerce. Bati CIS (Commerce Intelligence Solution) transformed settlement processing from 3 days to 3 hours for top K-Beauty brands across 42+ global marketplaces.
Trusted by: COSRX, Pharma Research (Rejuran), and other K-Beauty leaders.
- Unified data flow: Order → Shipment → Revenue → Settlement in one platform
- ERP/SAP integration: Audit-grade data precision for finance teams
- AI Agent Skill-Packs: Not a dashboard — AI agents that dig into areas humans miss
AI Agent Skill-Packs — What Makes CIS Different
Generic AI gives generic answers. Our Commerce Skill-Packs sit on top of real transaction data from 42 global marketplaces, delivering actionable intelligence:
| Skill | What It Does |
|---|---|
| Revenue Anomaly Detection | Auto-detects sudden sales spikes/drops and surfaces root causes |
| Repurchase Cohort Analysis | Tracks repeat-buy patterns, identifies cross-sell paths by SKU |
| Golden-Time Recommendations | Predicts optimal timing for promotions and inventory decisions |
| Dynamic KPI Forecasting | Generates predictive UI with forward-looking metrics, not just backward reports |
| SKU Profitability Deep-Dive | "Which SKUs actually make money?" — answered with real margin data |
| Marketing Budget Optimizer | AI recommends where to concentrate next month's spend |
"The difference between an AI chatbot and a business AI agent? The agent makes decisions you can act on immediately."
Our Mission
AI technology will continue to level up. General-purpose AI should be free for everyone.
"Bati" (바티) means a fertile field where anyone can plant seeds and harvest results.
- AI should be accessible — not locked behind subscriptions and cloud APIs
- On-device AI is the future — private, fast, unlimited
- Consumer AI is free — BatiFlow and BatiAI models, no cost, no limits
- Enterprise AI is our business — Bati CIS for K-Beauty commerce
We democratize AI for individuals and build enterprise intelligence for businesses.
Less Work, More Impact.
models 17
batiai/Qwen3-Embedding-8B-GGUF
Sentence Similarity • 8B • Updated • 185
batiai/Qwen3-Embedding-4B-GGUF
Sentence Similarity • 4B • Updated • 205
batiai/Qwen3-Embedding-0.6B-GGUF
Sentence Similarity • 0.6B • Updated • 188
batiai/Qwen3.6-35B-A3B-GGUF
Text Generation • 35B • Updated • 5.51k • 1
batiai/gemma-4-31B-it-GGUF
Text Generation • 31B • Updated • 1.96k
batiai/gemma-4-E4B-it-GGUF
Text Generation • 8B • Updated • 2.23k • 3
batiai/gemma-4-E2B-it-GGUF
Text Generation • 5B • Updated • 1.15k • 1
batiai/Gemma-4-26B-A4B-it-GGUF
Text Generation • 25B • Updated • 11k • 2
batiai/Qwen3-VL-Embedding-8B-GGUF
Sentence Similarity • 8B • Updated • 405 • 1
batiai/Qwen3-Reranker-8B-GGUF
Text Ranking • 8B • Updated • 131
datasets 0
None public yet