Pleias-RAG 1B

Overview

A 1.2B RAG-specialized model from PleIAs with built-in citation and grounding behavior. Beats most sub-4B small language models on HotPotQA.

When to pick this model

• You're deploying RAG on tight hardware budgets or edge devices
• You need clean citations and grounding from a small model
• You're handling structured Q&A where source attribution matters
• You want a defensible audit trail for regulated RAG deployments

VRAM requirements by quantization

Quantization	VRAM required
Q4_K_M (recommended)	0.8 GB
Q5_K_M	1 GB
Q8_0	1.5 GB
FP16 (no quantization)	2.5 GB

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Strengths

• Built-in citation and grounding in RAG responses
• Outperforms most small language models under 4B on HotPotQA
• Runs on lightweight hardware
• Apache 2.0

Limitations

• Context window of only ~2K
• No official Ollama tag
• Specialized for RAG — not a general chat model

Architecture & training

Architecture: Dense 1.2B · fine-tuned for RAG with built-in citations/grounding

Training: Based on Pleias 1.2B.

Quick start

# HuggingFace : PleIAs/Pleias-RAG-1B (GGUF : PleIAs/Pleias-RAG-1B-gguf)

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.