Instructions to use Nanthasit/sakthai-context-1.5b-merged-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Nanthasit/sakthai-context-1.5b-merged-v2 with Transformers:

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

pipe = pipeline("text-generation", model="Nanthasit/sakthai-context-1.5b-merged-v2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Nanthasit/sakthai-context-1.5b-merged-v2")
model = AutoModelForCausalLM.from_pretrained("Nanthasit/sakthai-context-1.5b-merged-v2", 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 Nanthasit/sakthai-context-1.5b-merged-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Nanthasit/sakthai-context-1.5b-merged-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Nanthasit/sakthai-context-1.5b-merged-v2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Nanthasit/sakthai-context-1.5b-merged-v2

SGLang

How to use Nanthasit/sakthai-context-1.5b-merged-v2 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 "Nanthasit/sakthai-context-1.5b-merged-v2" \
    --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": "Nanthasit/sakthai-context-1.5b-merged-v2",
		"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 "Nanthasit/sakthai-context-1.5b-merged-v2" \
        --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": "Nanthasit/sakthai-context-1.5b-merged-v2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Nanthasit/sakthai-context-1.5b-merged-v2 with Docker Model Runner:
```
docker model run hf.co/Nanthasit/sakthai-context-1.5b-merged-v2
```

SakThai Context 1.5B merged v2 🧠

Merged full-model tool-calling · Qwen2.5-1.5B · CPU-friendly · 3.1 GB safetensors

Part of the SakThai family — merged, full-parameter release of the 1.5B tool-calling line. Built on zero budget, runs on CPU, and remains one of the most accessible local tool-use models in the family.

Model Description

A merged, full-parameter Qwen2.5-1.5B model fine-tuned for structured tool-calling via <tools> XML schemas and <tool_call> JSON outputs. It is intended for single- and multi-tool use cases where you want a model that stays on-task without out-of-scope tool calls.

Parameter	Value
Architecture	Qwen2ForCausalLM (`qwen2`)
Parameters	~1.54 B
Hidden size	1,536
Layers	28
Attention heads	12 (GQA, 2 KV heads)
Intermediate size	8,960
Vocabulary	151,936
Context length	32,768 (32K)
RoPE theta	1,000,000
Base dtype	bfloat16
License	Apache 2.0
Size	~3.1 GB (safetensors)

How to Use

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Nanthasit/sakthai-context-1.5b-merged-v2"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cpu")

messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Find news about Tesla, then get its latest stock price."}
]

tools = [
    {"type": "function", "function": {
        "name": "search_web",
        "description": "Search recent news",
        "parameters": {"type": "object", "properties": {"query": {"type": "string"}}, "required": ["query"]}
    }},
    {"type": "function", "function": {
        "name": "get_stock_price",
        "description": "Get current stock price by ticker",
        "parameters": {"type": "object", "properties": {"ticker": {"type": "string"}}, "required": ["ticker"]}
    }}
]

prompt = tokenizer.apply_chat_template(messages, tools=tools, tokenize=False)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=1024, temperature=0.7, top_p=0.8)
print(tokenizer.decode(out[0][inputs.input_ids.shape[1]:], skip_special_tokens=False))

llama.cpp / GGUF note

This repo ships safetensors only. If you want CPU inference with llama.cpp / Ollama, use the GGUF companion:

# companion GGUF (if available in the family)
ollama run Nanthasit/sakthai-context-1.5b-merged

Benchmarks & Evaluation Status

SakThai Bench v2 (verified artifacts)

Metric	Score	Verified
Strict Accuracy	34.2	✅
Selection Accuracy	34.9	✅
Arguments Accuracy	44.2	✅

Source: .eval_results/sakthai-bench-v2.yaml in this repo, verified 2026-07-31.

Internal smoke test (cron eval, 2026-07-31)

Metric	Score	Verified
Multi-tool correctness	100%	✅
Valid JSON rate	100%	✅

Backend: llama.cpp Q4_K_M on CPU, 3 trials, multi-tool prompt (search_web + get_stock_price). Single-prompt smoke test, not a full benchmark.

Limitations

Merged full-parameter release; no PEFT/LoRA adapter.
Benchmarks are limited: bench-v2 scores show argument accuracy is still weak on this scale.
Strong <tool> schema requirement; tool-calling degrades without explicit schemas.
English-first training; non-English tool use is not validated.
32K context is config-supported but not stress-tested in this card.
Serverless Inference API may be unreliable for merged 3.1 GB safetensors on free tiers.
Internal benchmark is a single-prompt smoke test; treat 100% as presence-of-life, not full capability.
Evaluation artifacts are local/cron-generated; no third-party evaluation has been run on this merged artifact yet.

Citation

@misc{sakthai-context-1.5b-merged-v2,
  title  = {SakThai Context 1.5B merged v2},
  author = {Beer, Nanthasit},
  year   = {2026},
  url    = {https://huggingface.co/Nanthasit/sakthai-context-1.5b-merged-v2}
}
@misc{qwen2.5,
  title  = {Qwen2.5: A Party of Foundation Models},
  author = {Qwen Team},
  year   = {2024},
  url    = {https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct}
}

Sibling Models

Model	Size	Role	Downloads
context-1.5b-merged	3.1 GB	Flagship tool-calling merged	1,599
context-1.5b-merged-v2 (you are here)	3.1 GB	v2 merged	337
context-0.5b-merged	988 MB	Lightweight merged	1,370
context-7b-merged	15.2 GB	Full-power merged	744
context-7b-128k	recipe	128K long-context config	506
plus-1.5b	3.1 GB	Plus merged	0
plus-1.5b-lora	74 MB	Plus adapter	0

Downloads verified live 2026-07-31 via HF API.

Links

House of Sak · GitHub · All models · All datasets

License

Apache 2.0.

Evaluation & Verification

Benchmarks in this card are either directly read from repo artifacts (.eval_results/*.yaml) or referenced from the base model. Smoke-test metrics are single-prompt local runs, not third-party benchmarks, and are labeled verified: false where appropriate. If you independently re-run these evals, Beer asks that you share results back to the leaderboard Space.