Instructions to use AIJian/TrustSQL-4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AIJian/TrustSQL-4B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="AIJian/TrustSQL-4B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("AIJian/TrustSQL-4B")
model = AutoModelForCausalLM.from_pretrained("AIJian/TrustSQL-4B", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use AIJian/TrustSQL-4B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AIJian/TrustSQL-4B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AIJian/TrustSQL-4B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AIJian/TrustSQL-4B

SGLang

How to use AIJian/TrustSQL-4B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "AIJian/TrustSQL-4B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AIJian/TrustSQL-4B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "AIJian/TrustSQL-4B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AIJian/TrustSQL-4B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AIJian/TrustSQL-4B with Docker Model Runner:
```
docker model run hf.co/AIJian/TrustSQL-4B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

TRUST-SQL-4B

Overview

TrustSQL-4B is a fine-tuned Text-to-SQL model based on Qwen3-4B, introduced in TRUST-SQL: Tool-Integrated Multi-Turn Reinforcement Learning for Text-to-SQL over Unknown Schemas. The model is trained with multi-turn reinforcement learning and tool integration for Text-to-SQL over unknown database schemas.

Model details

Base model: Qwen/Qwen3-4B
Architecture: Qwen3ForCausalLM
Parameters: 4.0B
Hidden size: 2560
Layers: 36
Attention heads: 32 Q heads / 8 KV heads
Context length: 40,960 tokens
Precision: bfloat16

Models

Model	Base	Link
TrustSQL-4B	Qwen3-4B	AIJian/TrustSQL-4B
TrustSQL-8B	Qwen3-8B	AIJian/TrustSQL-8B

Training

TrustSQL follows a two-stage training pipeline: SFT warm-up followed by Phase-Aware GRPO optimization. The interaction protocol is Explore → Propose → Generate → Confirm.

Reported results

All results are reported under the Unknown Schema setting.

Benchmark	Greedy	Majority voting
BIRD-Dev	64.9	67.2
Spider-Test	82.8	85.0
Spider-DK	71.6	73.8
Spider-Syn	74.7	77.3
Spider-Realistic	79.9	82.5

Recommended inference setup

This model is designed for an orchestrator that exposes:

A schema exploration tool for tables, columns, keys, and value inspection.
A schema proposal channel that records verified tables and columns.
A SQL execution tool for candidate queries.
A final answer channel for the confirmed SQL.

Do not provide fabricated schema descriptions as if they were tool observations. The model is intended to ground schema decisions in the environment feedback.

Loading

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "AIJian/TrustSQL-4B"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True,
)

For prompts, tool schemas, evaluation scripts, and training details, see https://github.com/JaneEyre0530/TrustSQL.

Limitations

This checkpoint was trained and evaluated with SQLite-based benchmarks. Its behavior depends on a live, correctly configured tool environment and a finite interaction budget. Validate generated SQL before using it in any sensitive or write-enabled database.

Citation

@article{jian2026trustsql,
  title   = {TRUST-SQL: Tool-Integrated Multi-Turn Reinforcement Learning for Text-to-SQL over Unknown Schemas},
  author  = {Jian, Ai and Zhang, Xiaoyun and Du, Wanrou and Ruan, Jingqing and Pei, Jiangbo and Zhang, Weipeng and Zeng, Ke and Cai, Xunliang},
  journal = {arXiv preprint arXiv:2603.16448},
  year    = {2026}
}