TNH Agent (SmolLM3 3B) — Pure Thich Nhat Hanh Voice

A fine-tuned SmolLM3-3B model trained exclusively on Thich Nhat Hanh's teaching atoms. Embodied practice, interbeing, gentle firmness. Question in, TNH out.

This is the lightweight 3B branch of the TNH Agent. Same V6 training format as the 8B version, smaller footprint.

~4,100 atoms from 2 books (The Heart of Buddha's Teaching + Dharma Talks, 1,097 pages).

Quick Start

LM Studio / Ollama / llama.cpp: Download TNH_Agent-SmolLM3-3B-Q5_K_M.gguf and load it. No system prompt required.

Example:

User: I keep getting caught in anxious thinking. What do I do?

TNH Agent: When you notice you are caught, that noticing is already the beginning of freedom. You don't have to fight the anxious thinking — you have to recognize it. "I see you, anxiety." Recognition is not suppression. Breathe in, and with the in-breath, say hello to what is there. Breathe out, and allow it to be held in the warmth of your awareness. This is the practice of mindfulness — not eliminating, but embracing.

User: What is interbeing?

TNH Agent: Look at a piece of paper. In it you can see the cloud — without the cloud, there is no rain; without rain, the tree cannot grow; without the tree, there is no paper. The cloud is in the paper. The logger who cut the tree is in the paper. The wheat that fed the logger is in the paper. Everything inter-is. Nothing can exist by itself alone.

What Makes This Different

  • Pure TNH, nothing else. Trained only on Thich Nhat Hanh's atoms.
  • Embodied practice. Breathing, walking, touching the earth — practice is present-moment return.
  • Soft firmness. Not harsh, not soft — grounded in the dharma without aggression.
  • Interbeing alive. The model carries his specific way of seeing connection.
  • 3B footprint. Runs on modest hardware — ~2 GB with Q5_K_M.

Training Details

Setting Value
Base model HuggingFaceTB/SmolLM3-3B-Base
Method A-LoRA V6 (QDoRA + rsLoRA)
Rank 32, alpha 32
Training data TNH V6 atoms (from ~4,100 raw, filtered for complete question+teaching pairs)
Source books The Heart of Buddha's Teaching, Dharma Talks
Format V6 — question + concept arrows (input) → pure teaching passage (output)
Epochs 1
Quantization Q5_K_M GGUF (~2 GB)

The A-LoRA Methodology

A-LoRA (Atom LoRA) fine-tunes on complete reasoning atoms — indivisible multi-component teaching moves extracted from human expert traces. Each atom contains transformation (before/after), concept relations, anchoring quotes, and teacher method. Training on whole atoms produces models that move like teachers, not just sound like them.

The Teacher

Thich Nhat Hanh (1926–2022). Engaged Buddhism. Interbeing. The breath as the bridge between mind and body. Walking meditation as peace in every step. His teaching is not conceptual — it is a practice you can do with your next breath.

Style element What it means
Embodied practice Instruction grounded in the body — breath, walking, eating
Interbeing Everything exists in relationship; nothing is separate
Soft firmness Gentle but uncompromising in pointing to what is
Present-moment return Every teaching is an invitation back to now

3B vs 8B

The 8B TNH Agent carries more nuance and handles multi-step practice guidance better. This 3B branch preserves the core TNH movement for memory-constrained environments.

Files

File Size Use
TNH_Agent-SmolLM3-3B-Q5_K_M.gguf ~2 GB Recommended

Related Models


ellam sivamayam — Everything is Shiva's expression.

எல்லாம் சிவமயம்

Downloads last month
46
GGUF
Model size
3B params
Architecture
smollm3
Hardware compatibility
Log In to add your hardware

5-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Sathman/TNH-Agent-SmolLM3-3B-GGUF

Adapter
(17)
this model

Collection including Sathman/TNH-Agent-SmolLM3-3B-GGUF