Qwen3-Coder-Next · 80B A3B
Qwen3-Coder-Next is an open-weight text-generation model from Qwen/Qwen3-Coder-Next, licensed under apache-2.0. It is designed specifically for coding agents and local development. The model has 80B total parameters with 3B activated parameters, a 262,144-token native context length, and a hybrid architecture combining Gated DeltaNet, Gated Attention, and Mixture-of-Experts layers. It supports only non-thinking mode and does not generate thinking blocks.
What do you want to do?
Choose a task. Get the software, files, settings, and steps to run it.
Chat locally with Qwen3-Coder-Next
Chat privately with Qwen3-Coder-Next in the Ollama app on Windows or Mac, or its interactive Terminal on Linux. Choose official Q4, higher-precision Q8, or an uncensored package, then start a conversation with a practical 4K context.
Build a local AI agent with Qwen3-Coder-Next
Connect Qwen3-Coder-Next to Cline for a local AI agent that can work with files in your workspace. The example compares two product briefs, shows its file tools in action, and saves a cited result for you to inspect.
Build a local coding assistant with Qwen3-Coder-Next
Connect Qwen3-Coder-Next to Cline in VS Code for a private local coding assistant. The example changes a supplied calculator project, then checks the diff and runs independent tests.
Review code privately with Qwen3-Coder-Next on rented GPUs
Rent two NVIDIA GPUs to serve Qwen3-Coder-Next privately through vLLM, connect Cline in your local VS Code, and review a local repository and inspect its suggested fix.
Open a guide for its software, hardware requirements, and setup steps.
At a glance
- Parameters
- 80B total / 3B active
- Architecture
- Hybrid attention MoE
- Context length
- 262,144 native
- License
- Apache 2.0
- Disk space
- 52 GB
- Software
- Ollama, vLLM, SGLang
Will it run on your machine?
Save your machine to see a personalized rating and its reasoning.
Add my hardwareBest for
- Coding agents
- Local development
- Long-horizon reasoning
- Complex tool usage
- Integration with CLI/IDE platforms
Tradeoffs
- Supports only non-thinking mode
- Does not generate thinking blocks
- No explicit license, compatibility, performance, or numerical hardware requirements are provided beyond the apache-2.0 license
Ways to run it
vLLM · Linux
Repository-specific example documented in the publisher model card; confirm dependencies and platform in the source.
vllm serve Qwen/Qwen3-Coder-Next --port 8000 --tensor-parallel-size 2 --enable-auto-tool-choice --tool-call-parser qwen3_coderTransformers · See official guide
Repository-specific example documented in the publisher model card; confirm dependencies and platform in the source.
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "Qwen/Qwen3-Coder-Next"
# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype="auto",
device_map="auto"
)
# prepare the model input
prompt = "Write a quick sort algorithm."
messages = [
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
# conduct text completion
generated_ids = model.generate(
**model_inputs,
max_new_tokens=65536
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()
content = tokenizer.decode(output_ids, skip_special_tokens=True)
print("content:", content)Explore its uses
Keep exploring
Qwen2.5 Coder · 7B
Code generation, repair, and reasoning in a compact open-weight coding model.
Mistral Small 4 · 119B
Mistral Small 4 119B A6B is an Apache-2.0 licensed hybrid model from mistralai that combines general instruction, reasoning, and Devstral-style capabilities in one model. It uses a Mixture-of-Experts architecture with 119B total parameters, 6.5B activated per token, 128 experts with 4 active, and a 256k context length. It supports multimodal input (text and image) with text output, multilingual support across dozens of languages, system prompts, function calling, JSON output, and configurable reasoning effort. The model card also mentions optional efficiency resources: an eagle head for speculative decoding and an NVFP4 checkpoint. It is tagged for vLLM and offered in safetensors format.
Mellum2.1 Thinking
A 12B-total, 2.5B-active open-weight thinking model for coding and tool use. Start with JetBrains’ official GGUF on a suitable desktop, or serve the BF16 checkpoint with vLLM on a larger GPU.