Expired50% confidenceReleaseExact time
NVIDIA released an NVFP4 quantized model of Qwen3 235B on Hugging Face, compressing the model from approximately 470 GB of VRAM in BF16 to approximately 156 GB, enabling it to run on a single-node cluster of 8 H100 GPUs.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/19/2026
First Seen
Valid until: 10/17/2026(expired)
Sources
Related Claims
UnverifiedAn NVIDIA H100 single card has 80 GB of VRAM, and an 8-card node with 640 GB is still insufficient to load the full weights of Qwen3.8.75% similarUnverifiedThe NVIDIA A100 has a memory bandwidth of 2 TB/s; loading a 280 GB FP32 model takes approximately 140 ms, meaning weight loading alone limits generation to a maximum of approximately 7 tokens per second.73% similarVerifiedA single NVIDIA A100 card has 80GB HBM2e memory and 312 TFLOPS of FP16 computing power72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/561864API
curl https://kongchang.com/api/v1/knowledge/claims/561864MCP
get_claim(id=561864)