Qwen3.8-Max

Qwen3.8-Max

This repository contains the open-weight version of Qwen3.8-Max, officially released as Qwen3.8-2.4T-A95B.

Model Overview

• Architecture: Sparse Mixture-of-Experts (MoE)
• Total Parameters: 2.4 Trillion
• Active Parameters: 95 Billion per token
• Experts: 512 total (10 routed + 1 shared activated per token)
• Context Length: 262,144 tokens natively (extensible up to ~1M tokens)
• Type: Causal Language Model
• Modalities: Text → Text (open weights version)
• License: Qwen3.8-Max License

Note: The official hosted version of Qwen3.8-Max (available via Qwen Cloud / Alibaba Cloud) includes additional features such as native vision/video input, non-thinking mode, and built-in tools. The open-weight version focuses on text generation with strong reasoning capabilities.

Model Details

Usage

With Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Vili-Elvis01/Qwen3.8-Max"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

prompt = "Explain quantum computing in simple terms."
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)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=512
)

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)