The community can now stop refreshing Hugging Face every 11 seconds. After 14 months of "it's almost ready," the open-weight model you've been dreaming about is now available for download, local inference, and mild existential dread.
Compared against leading models and your current coping mechanism
Choose your preferred method of suffering
Original BF16 weights — 60GB of pure joy
Q4_K_M — because your GPU has dreams
$0.0004 per 1K tokens — cheaper than therapy
For the true believers. 284 seeders and counting.
A timeline of promises, delays, and one very confused project manager
Because you've waited long enough to not need more than three
# You made it this far. You deserve this model. from transformers import AutoModelForCausalLM, AutoTokenizer model_name = "Qwen/Qwen3.7-30B-Open" tokenizer = AutoTokenizer.from_pretrained(model_name) model = AutoModelForCausalLM.from_pretrained( model_name, device_map="auto", torch_dtype="bfloat16", trust_remote_code=True, ) messages = [{"role": "user", "content": "Explain why this took 14 months"}] text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) outputs = model.generate(*tokenizer.__call__(text).to(model.device), max_new_tokens=512) print(tokenizer.decode(outputs[0], skip_special_tokens=True)) # Output: "Let me start by saying that our team worked tirelessly..."
Real quotes from people who stayed up all night waiting for this
Suggested VRAM setups, ranked by how much you regret buying GPUs
RTX 4090 · Q4 quant · 8K context · pray to the silicon gods
Dual 3090s · Q6 quant · 32K context · you're a legend
4× 3090s · Full BF16 · 128K context · your electric bill weeps
Intel i9 · GGUF Q2 · 256 tokens · by the time it finishes, AGI will have replaced you