vLLM on X: "🚀 Qwen3.6-27B-NVFP4 is inference ready with vLLM on NVIDIA Blackwell GPUs. This checkpoint is optimized for Blackwell and reduces GPU memory requirements by ~2.5x for local AI with open-source models. 🧠 27B params, Hybrid Attention 📊 NVFP4 evals: 86.3 on MMLU Pro, 85.5 on" / X

Qwen3.6-27B-NVFP4 is now inference-ready on NVIDIA Blackwell GPUs, optimizing GPU memory usage by ~2.5x for local AI. With 27 billion parameters and hybrid attention, it achieves 86.3 on MMLU Pro and 85.5 on GPQA Diamond, exclusively supported by vLLM.

Key Points

• Optimized for NVIDIA Blackwell GPUs, enhancing local AI performance.
• Reduces GPU memory requirements by approximately 2.5x.
• Achieves 86.3 on MMLU Pro and 85.5 on GPQA Diamond benchmarks.
• Exclusively supported on vLLM as the runtime engine.
• Available on Hugging Face for easy access to open-source models.