TX-12G

Enhanced reasoning and code generation. Runs on 12GB RAM.

TX-12G is TARX's mid-tier model, offering significantly improved reasoning and code generation while remaining accessible to users with 12GB+ RAM.

Model Details

Property	Value
Parameters	12B
Quantization	Optimized mixed precision
RAM Required	12 GB minimum
GPU VRAM	8 GB+ recommended
Context Length	16,384 tokens
License	Apache 2.0

Capabilities

• ✅ Everything TX-8G does, plus:
• ✅ Complex multi-step reasoning
• ✅ Advanced code generation & debugging
• ✅ Nuanced writing with style matching
• ✅ Technical documentation
• ✅ Data analysis & interpretation

When to Use TX-12G vs TX-8G

Use Case	TX-8G	TX-12G
Quick questions	✅	Overkill
Email drafting	✅	✅
Simple code	✅	✅
Complex debugging	⚠️	✅
Multi-file refactoring	❌	✅
Technical writing	⚠️	✅
Research synthesis	⚠️	✅

Usage

With TARX Desktop

Settings → Model → TX-12G

Model downloads automatically on first use (~8GB download).

With Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "Tarxxxxxx/TX-12G",
    device_map="auto",
    torch_dtype="auto"
)
tokenizer = AutoTokenizer.from_pretrained("Tarxxxxxx/TX-12G")

With llama.cpp

wget https://huggingface.co/Tarxxxxxx/TX-12G/resolve/main/tx-12g.Q6_K.gguf
./main -m tx-12g.Q6_K.gguf -p "Debug this Python function:" -n 512

Hardware Requirements

Hardware	Performance
Apple M1 Pro/Max (16GB+)	⭐⭐⭐⭐⭐ Excellent
Apple M2/M3 (16GB+)	⭐⭐⭐⭐⭐ Excellent
NVIDIA RTX 3080+	⭐⭐⭐⭐⭐ Excellent
Intel i7 + 32GB RAM	⭐⭐⭐⭐ Good
AMD Ryzen 7 + 32GB	⭐⭐⭐⭐ Good

Links

• TARX Website
• Documentation
• Download

---

Built by TARX | tarx.com