JetBrains-Research/Qwen3-30B-A3B-am

Modified Qwen3-30B-A3B

This is a modified version of Qwen/Qwen3-30B-A3B with an updated tokenizer configuration for improved agentic mode with function calling.

Modifications

The tokenizer configuration has been updated with:

• Enhanced chat template with {% generation %} tags for better tool mask handling
• Added extra_special_tokens field
• Improved compatibility with agentic workflows

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("JetBrains-Research/Qwen3-30B-A3B-am")
tokenizer = AutoTokenizer.from_pretrained("JetBrains-Research/Qwen3-30B-A3B-am")

# Use with function calling / agentic mode
messages = [
    {"role": "user", "content": "Your prompt here"}
]

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

Original Model

This model is based on Qwen/Qwen3-30B-A3B. Please refer to the original model card for more information about capabilities and limitations.