A standalone PowerShell module provides the fastest route to local installation.
Follow the guidelines below to continue.
Everything happens automatically, including the heavy cloud asset download.
The setup file includes a feature that instantly optimizes all configurations.
The Qwen3.5-397B-A17B-NVFP4 model represents a major leap in large language model efficiency, combining a 397‑billion parameter architecture with the ultra‑low‑precision NVFP4 data type.
By leveraging NVFP4 quantization, the model achieves a dramatic reduction in memory footprint while preserving near‑full‑precision performance, making it ideal for deployment on consumer‑grade GPUs.
Benchmarks show that the model delivers sub‑50 ms inference latency and a throughput of over 200 tokens per second on standard hardware, outperforming previous 400B‑scale models.
Its training pipeline incorporates a novel mixture‑of‑experts routing scheme that balances load across the A17B accelerator cluster, resulting in stable convergence and robust multilingual capabilities.
The integrated
| Model | Parameters | Precision | Latency (ms) | Throughput (tokens/s) |
|---|---|---|---|---|
| Qwen3.5-397B-A17B-NVFP4 | 397B | NVFP4 | <50 | >200 |
provides a quick comparison with competing models, highlighting parameter count, precision, latency, and throughput in a concise format.