Mungert/SmolLM3-3B-GGUF

SmolLM3-3B GGUF Models

Model Generation Details

This model was generated using llama.cpp at commit 670e1360.

---

Click here to get info on choosing the right GGUF model format

---

SmolLM3

Table of Contents

1. Model Summary
2. How to use
3. Evaluation
4. Training
5. Limitations
6. License

Model Summary

SmolLM3 is a 3B parameter language model designed to push the boundaries of small models. It supports dual mode reasoning, 6 languages and long context. SmolLM3 is a fully open model that offers strong performance at the 3B–4B scale.

The model is a decoder-only transformer using GQA and NoPE (with 3:1 ratio), it was pretrained on 11.2T tokens with a staged curriculum of web, code, math and reasoning data. Post-training included midtraining on 140B reasoning tokens followed by supervised fine-tuning and alignment via Anchored Preference Optimization (APO).

Key features

• Instruct model optimized for hybrid reasoning
• Fully open model: open weights + full training details including public data mixture and training configs
• Long context: Trained on 64k context and suppots up to 128k tokens using YARN extrapolation
• Multilingual: 6 natively supported (English, French, Spanish, German, Italian, and Portuguese)

For more details refer to our blog post: https://hf.co/blog/smollm3

How to use

The modeling code for SmolLM3 is available in transformers v4.53.0, so make sure to upgrade your transformers version. You can also load the model with the latest vllm which uses transformers as a backend.

pip install -U transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "HuggingFaceTB/SmolLM3-3B"
device = "cuda"  # for GPU usage or "cpu" for CPU usage

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
).to(device)

# prepare the model input
prompt = "Give me a brief explanation of gravity in simple terms."
messages_think = [
    {"role": "user", "content": prompt}
]

text = tokenizer.apply_chat_template(
    messages_think,
    tokenize=False,
    add_generation_prompt=True,
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

# Generate the output
generated_ids = model.generate(**model_inputs, max_new_tokens=32768)

# Get and decode the output
output_ids = generated_ids[0][len(model_inputs.input_ids[0]) :]
print(tokenizer.decode(output_ids, skip_special_tokens=True))

We recommend setting temperature=0.6 and top_p=0.95 in the sampling parameters.

Long context processing

The current config.json is set for context length up to 65,536 tokens. To handle longer inputs (128k or 256k), we utilize YaRN you can change the max_position_embeddings and rope_scaling` to:

{
  ...,
  "rope_scaling": {
    "factor": 2.0, #2x65536=131 072 
    "original_max_position_embeddings": 65536,
    "type": "yarn"
  }
}

Enabling and Disabling Extended Thinking Mode

We enable extended thinking by default, so the example above generates the output with a reasoning trace. For choosing between enabling, you can provide the /think and /no_think flags through the system prompt as shown in the snippet below for extended thinking disabled. The code for generating the response with extended thinking would be the same except that the system prompt should have /think instead of /no_think.

prompt = "Give me a brief explanation of gravity in simple terms."
messages = [
    {"role": "system", "content": "/no_think"},
    {"role": "user", "content": prompt}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)

We also provide the option of specifying the whether to use extended thinking through the enable_thinking kwarg as in the example below. You do not need to set the /no_think or /think flags through the system prompt if using the kwarg, but keep in mind that the flag in the system prompt overwrites the setting in the kwarg.

prompt = "Give me a brief explanation of gravity in simple terms."
messages = [
    {"role": "user", "content": prompt}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=False
)

Agentic Usage

SmolLM3 supports tool calling! Just pass your list of tools:

• Under the argument xml_tools for standard tool-calling: these tools will be called as JSON blobs within XML tags, like <tool_call>{"name": "get_weather", "arguments": {"city": "Copenhagen"}}</tool_call>
• Or under python_tools: then the model will call tools like python functions in a <code> snippet, like <code>get_weather(city="Copenhagen")</code>

from transformers import AutoModelForCausalLM, AutoTokenizer

checkpoint = "HuggingFaceTB/SmolLM3-3B"

tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForCausalLM.from_pretrained(checkpoint)

tools = [
    {
        "name": "get_weather",
        "description": "Get the weather in a city",
        "parameters": {"type": "object", "properties": {"city": {"type": "string", "description": "The city to get the weather for"}}}}
]

messages = [
    {
        "role": "user",
        "content": "Hello! How is the weather today in Copenhagen?"
    }
]

inputs = tokenizer.apply_chat_template(
    messages,
    enable_thinking=False, # True works as well, your choice!
    xml_tools=tools,
    add_generation_prompt=True,
    tokenize=True,
    return_tensors="pt"
)

outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))

Using Custom System Instructions.

You can specify custom instruction through the system prompt while controlling whether to use extended thinking. For example, the snippet below shows how to make the model speak like a pirate while enabling extended thinking.

prompt = "Give me a brief explanation of gravity in simple terms."
messages = [
    {"role": "system", "content": "Speak like a pirate./think"},
    {"role": "user", "content": prompt}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)

For local inference, you can use llama.cpp, ONNX, MLX and MLC. You can find quantized checkpoints in this collection (https://huggingface.co/collections/HuggingFaceTB/smollm3-686d33c1fdffe8e635317e23)

vLLM and SGLang

You can use vLLM and SGLang to deploy the model in an API compatible with OpenAI format.

SGLang

python -m sglang.launch_server --model-path HuggingFaceTB/SmolLM3-3B

vLLM

vllm serve HuggingFaceTB/SmolLM3-3B --enable-auto-tool-choice --tool-call-parser=hermes

Setting chat_template_kwargs

You can specify chat_template_kwargs such as enable_thinking to a deployed model by passing the chat_template_kwargs parameter in the API request.

curl http://localhost:8000/v1/chat/completions -H "Content-Type: application/json" -d '{
  "model": "HuggingFaceTB/SmolLM3-3B",
  "messages": [
    {"role": "user", "content": "Give me a brief explanation of gravity in simple terms."}
  ],
  "temperature": 0.6,
  "top_p": 0.95,
  "max_tokens": 16384,
  "chat_template_kwargs": {"enable_thinking": false}
}'

Evaluation

In this section, we report the evaluation results of SmolLM3 model. All evaluations are zero-shot unless stated otherwise, and we use lighteval to run them.

We highlight the best score in bold and underline the second-best score.

Instruction Model

No Extended Thinking

Evaluation results of non reasoning models and reasoning models in no thinking mode. We highlight the best and second-best scores in bold.

Category	Metric	SmoLLM3-3B	Qwen2.5-3B	Llama3.1-3B	Qwen3-1.7B	Qwen3-4B
High school math competition	AIME 2025	9.3	2.9	0.3	8.0	17.1
Math problem-solving	GSM-Plus	72.8	74.1	59.2	68.3	82.1
Competitive programming	LiveCodeBench v4	15.2	10.5	3.4	15.0	24.9
Graduate-level reasoning	GPQA Diamond	35.7	32.2	29.4	31.8	44.4
Instruction following	IFEval	76.7	65.6	71.6	74.0	68.9
Alignment	MixEval Hard	26.9	27.6	24.9	24.3	31.6
Tool Calling	BFCL	92.3	-	92.3 *	89.5	95.0
Multilingual Q&A	Global MMLU	53.5	50.54	46.8	49.5	65.1

(*): this is a tool calling finetune

Extended Thinking

Evaluation results in reasoning mode for SmolLM3 and Qwen3 models:

Category	Metric	SmoLLM3-3B	Qwen3-1.7B	Qwen3-4B
High school math competition	AIME 2025	36.7	30.7	58.8
Math problem-solving	GSM-Plus	83.4	79.4	88.2
Competitive programming	LiveCodeBench v4	30.0	34.4	52.9
Graduate-level reasoning	GPQA Diamond	41.7	39.9	55.3
Instruction following	IFEval	71.2	74.2	85.4
Alignment	MixEval Hard	30.8	33.9	38.0
Tool Calling	BFCL	88.8	88.8	95.5
Multilingual Q&A	Global MMLU	64.1	62.3	73.3

Base Pre-Trained Model

English benchmarks

Note: All evaluations are zero-shot unless stated otherwise. For Ruler 64k evaluation, we apply YaRN to the Qwen models with 32k context to extrapolate the context length.

Category	Metric	SmolLM3-3B	Qwen2.5-3B	Llama3-3.2B	Qwen3-1.7B-Base	Qwen3-4B-Base
Reasoning & Commonsense	HellaSwag	76.15	74.19	75.52	60.52	74.37
	ARC-CF (Average)	65.61	59.81	58.58	55.88	62.11
	Winogrande	58.88	61.41	58.72	57.06	59.59
	CommonsenseQA	55.28	49.14	60.60	48.98	52.99
Knowledge & Understanding	MMLU-CF (Average)	44.13	42.93	41.32	39.11	47.65
	MMLU Pro CF	19.61	16.66	16.42	18.04	24.92
	MMLU Pro MCF	32.70	31.32	25.07	30.39	41.07
	PIQA	78.89	78.35	78.51	75.35	77.58
	OpenBookQA	40.60	40.20	42.00	36.40	42.40
	BoolQ	78.99	73.61	75.33	74.46	74.28
Math & Code
Coding & math	HumanEval+	30.48	34.14	25.00	43.29	54.87
	MBPP+	52.91	52.11	38.88	59.25	63.75
	MATH (4-shot)	46.10	40.10	7.44	41.64	51.20
	GSM8k (5-shot)	67.63	70.13	25.92	65.88	74.14
Long context
	Ruler 32k	76.35	75.93	77.58	70.63	83.98
	Ruler 64k	67.85	64.90	72.93	57.18	60.29
	Ruler 128k	61.03	62.23	71.30	43.03	47.23

Multilingual benchmarks

Category	Metric	SmolLM3 3B Base	Qwen2.5-3B	Llama3.2 3B	Qwen3 1.7B Base	Qwen3 4B Base
Main supported languages
French	MLMM Hellaswag	63.94	57.47	57.66	51.26	61.00
	Belebele	51.00	51.55	49.22	49.44	55.00
	Global MMLU (CF)	38.37	34.22	33.71	34.94	41.80
	Flores-200 (5-shot)	62.85	61.38	62.89	58.68	65.76
Spanish	MLMM Hellaswag	65.85	58.25	59.39	52.40	61.85
	Belebele	47.00	48.88	47.00	47.56	50.33
	Global MMLU (CF)	38.51	35.84	35.60	34.79	41.22
	Flores-200 (5-shot)	48.25	50.00	44.45	46.93	50.16
German	MLMM Hellaswag	59.56	49.99	53.19	46.10	56.43
	Belebele	48.44	47.88	46.22	48.00	53.44
	Global MMLU (CF)	35.10	33.19	32.60	32.73	38.70
	Flores-200 (5-shot)	56.60	50.63	54.95	52.58	50.48
Italian	MLMM Hellaswag	62.49	53.21	54.96	48.72	58.76
	Belebele	46.44	44.77	43.88	44.00	48.78
	Global MMLU (CF)	36.99	33.91	32.79	35.37	39.26
	Flores-200 (5-shot)	52.65	54.87	48.83	48.37	49.11
Portuguese	MLMM Hellaswag	63.22	57.38	56.84	50.73	59.89
	Belebele	47.67	49.22	45.00	44.00	50.00
	Global MMLU (CF)	36.88	34.72	33.05	35.26	40.66
	Flores-200 (5-shot)	60.93	57.68	54.28	56.58	63.43

The model has also been trained on Arabic (standard), Chinese and Russian data, but has seen fewer tokens in these languages compared to the 6 above. We report the performance on these langages for information.

Category	Metric	SmolLM3 3B Base	Qwen2.5-3B	Llama3.2 3B	Qwen3 1.7B Base	Qwen3 4B Base
Other supported languages
Arabic	Belebele	40.22	44.22	45.33	42.33	51.78
	Global MMLU (CF)	28.57	28.81	27.67	29.37	31.85
	Flores-200 (5-shot)	40.22	39.44	44.43	35.82	39.76
Chinese	Belebele	43.78	44.56	49.56	48.78	53.22
	Global MMLU (CF)	36.16	33.79	39.57	38.56	44.55
	Flores-200 (5-shot)	29.17	33.21	31.89	25.70	32.50
Russian	Belebele	47.44	45.89	47.44	45.22	51.44
	Global MMLU (CF)	36.51	32.47	34.52	34.83	38.80
	Flores-200 (5-shot)	47.13	48.74	50.74	54.70	60.53

Training

Model

• Architecture: Transformer decoder
• Pretraining tokens: 11T
• Precision: bfloat16

Software & hardware

• GPUs: 384 H100
• Training Framework: nanotron
• Data processing framework: datatrove
• Evaluation framework: lighteval
• Post-training Framework: TRL

Open resources

Here is an infographic with all the training details

• The datasets used for pretraining can be found in this collection and those used in mid-training and post-training will be uploaded later
• The training and evaluation configs and code can be found in the huggingface/smollm repository.

Limitations

SmolLM3 can produce text on a variety of topics, but the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data. These models should be used as assistive tools rather than definitive sources of information. Users should always verify important information and critically evaluate any generated content.

License

Apache 2.0

---

🚀 If you find these models useful

Help me test my AI-Powered Quantum Network Monitor Assistant with quantum-ready security checks:

👉 Quantum Network Monitor

The full Open Source Code for the Quantum Network Monitor Service available at my github repos ( repos with NetworkMonitor in the name) : Source Code Quantum Network Monitor. You will also find the code I use to quantize the models if you want to do it yourself GGUFModelBuilder

💬 How to test:
Choose an AI assistant type:

• TurboLLM (GPT-4.1-mini)
• HugLLM (Hugginface Open-source models)
• TestLLM (Experimental CPU-only)

What I’m Testing

I’m pushing the limits of small open-source models for AI network monitoring, specifically:

• Function calling against live network services
• How small can a model go while still handling:
  • Automated Nmap security scans
  • Quantum-readiness checks
  • Network Monitoring tasks

🟡 TestLLM – Current experimental model (llama.cpp on 2 CPU threads on huggingface docker space):

• ✅ Zero-configuration setup
• ⏳ 30s load time (slow inference but no API costs) . No token limited as the cost is low.
• 🔧 Help wanted! If you’re into edge-device AI, let’s collaborate!

Other Assistants

🟢 TurboLLM – Uses gpt-4.1-mini :

• **It performs very well but unfortunatly OpenAI charges per token. For this reason tokens usage is limited.
• Create custom cmd processors to run .net code on Quantum Network Monitor Agents
• Real-time network diagnostics and monitoring
• Security Audits
• Penetration testing (Nmap/Metasploit)

🔵 HugLLM – Latest Open-source models:

• 🌐 Runs on Hugging Face Inference API. Performs pretty well using the lastest models hosted on Novita.

💡 Example commands you could test:

1. "Give me info on my websites SSL certificate"
2. "Check if my server is using quantum safe encyption for communication"
3. "Run a comprehensive security audit on my server"
4. '"Create a cmd processor to .. (what ever you want)" Note you need to install a Quantum Network Monitor Agent to run the .net code on. This is a very flexible and powerful feature. Use with caution!

Final Word

I fund the servers used to create these model files, run the Quantum Network Monitor service, and pay for inference from Novita and OpenAI—all out of my own pocket. All the code behind the model creation and the Quantum Network Monitor project is open source. Feel free to use whatever you find helpful.

If you appreciate the work, please consider buying me a coffee ☕. Your support helps cover service costs and allows me to raise token limits for everyone.

I'm also open to job opportunities or sponsorship.

Thank you! 😊