Instructions to use modrill/Qwen3-4B-Base-ThinkCode-A-U025 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use modrill/Qwen3-4B-Base-ThinkCode-A-U025 with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-4B-Base")
model = PeftModel.from_pretrained(base_model, "modrill/Qwen3-4B-Base-ThinkCode-A-U025")

Transformers

How to use modrill/Qwen3-4B-Base-ThinkCode-A-U025 with Transformers:

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

pipe = pipeline("text-generation", model="modrill/Qwen3-4B-Base-ThinkCode-A-U025")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("modrill/Qwen3-4B-Base-ThinkCode-A-U025", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use modrill/Qwen3-4B-Base-ThinkCode-A-U025 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "modrill/Qwen3-4B-Base-ThinkCode-A-U025"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "modrill/Qwen3-4B-Base-ThinkCode-A-U025",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/modrill/Qwen3-4B-Base-ThinkCode-A-U025

SGLang

How to use modrill/Qwen3-4B-Base-ThinkCode-A-U025 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 "modrill/Qwen3-4B-Base-ThinkCode-A-U025" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "modrill/Qwen3-4B-Base-ThinkCode-A-U025",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "modrill/Qwen3-4B-Base-ThinkCode-A-U025" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "modrill/Qwen3-4B-Base-ThinkCode-A-U025",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use modrill/Qwen3-4B-Base-ThinkCode-A-U025 with Docker Model Runner:
```
docker model run hf.co/modrill/Qwen3-4B-Base-ThinkCode-A-U025
```

Qwen3-4B-Base-ThinkCode-A-U025

File size: 1,815 Bytes

4cd3552

{"arm_id":"A-U025","base_model":"Qwen/Qwen3-4B-Base","base_revision":"906bfd4b4dc7f14ee4320094d8b41684abff8539","fixed_prompt_fp32_quantization_isolated_logits":[{"max_abs":0.13026046752929688,"mean_abs":0.01463189348578453,"p99_abs":0.04955993592739105,"prompt_index":0,"top10_overlap":10,"top1_equal":true},{"max_abs":0.11943221092224121,"mean_abs":0.030642252415418625,"p99_abs":0.07536689192056656,"prompt_index":1,"top10_overlap":9,"top1_equal":true},{"max_abs":0.09069585800170898,"mean_abs":0.02466517873108387,"p99_abs":0.057938352227211,"prompt_index":2,"top10_overlap":10,"top1_equal":true}],"logit_gate":{"max_abs_le":0.25,"mean_abs_le":0.05,"top10_overlap_min":9,"top1_all_equal":true},"phase_a_canary_status":"PASS","representative_bf16_weight_parity":[{"bf16_exact_samples":6,"logical_module":"model.layers.0.self_attn.q_proj","max_abs":0.0,"physical_weight":"model.layers.0.self_attn.q_proj.weight","sample_count":6},{"bf16_exact_samples":6,"logical_module":"model.layers.23.mlp.down_proj","max_abs":0.0,"physical_weight":"model.layers.23.mlp.down_proj.weight","sample_count":6},{"bf16_exact_samples":6,"logical_module":"model.layers.24.mlp.up_proj","max_abs":0.0,"physical_weight":"model.layers.24.mlp.up_proj.weight","sample_count":6},{"bf16_exact_samples":6,"logical_module":"model.layers.35.self_attn.o_proj","max_abs":0.0,"physical_weight":"model.layers.35.self_attn.o_proj.weight","sample_count":6},{"bf16_exact_samples":6,"logical_module":"lm_head","max_abs":0.0,"physical_weight":"model.embed_tokens.weight","sample_count":6}],"schema":"HF_PEFT_ADAPTER_VALIDATION_V1","status":"PASS","tensor_contract":{"base_layer_tensor_absent":true,"body_scale":0.25,"lm_head_scale":0.25,"peft_tied_adapter_storage_verified":true,"source_scaling_all_exact":true},"validated_at_utc":"2026-08-22T18:05:45Z"}