Nvidia said in a newly published article that Alibaba’s latest preview model, Qwen3.8-Flash-Next, is now supported on the Nvidia GB300 NVL72 platform. The company said the model has 176 billion total parameters, with about 6 billion activated per token. It natively supports a 262,000-token context window and can be extended to 1 million tokens through YaRN.
According to Nvidia, the model is aimed at long-context agent use cases including intelligent coding, document processing, and tool calling. Nvidia also said Qwen3.8-Flash-Next uses a hybrid architecture that combines Gated DeltaNet, or GDN, with Qwen Sparse Attention, or QSA, to cut compute demand and KV cache overhead in long-context workloads.
In Nvidia’s tests on GB300 NVL72, single-GPU throughput exceeded 16,000 tokens per second, while single-user throughput topped 200 tokens per second. Nvidia added that the model supports inference frameworks including SGLang, vLLM, and TensorRT-LLM.
Nvidia said in a published article that Alibaba’s latest preview model, Qwen3.8-Flash-Next, has gained support on the Nvidia GB300 NVL72 platform.
Nvidia said the model has 176 billion total parameters, with roughly 6 billion parameters activated per token. It natively supports a 262,000-token context window and can be extended to 1 million tokens through YaRN. The model is mainly aimed at long-context agent applications, including intelligent coding, document processing, and tool calling.
Nvidia also said Qwen3.8-Flash-Next uses a hybrid architecture combining Gated DeltaNet (GDN) and Qwen Sparse Attention (QSA) to reduce compute demand and KV cache overhead in long-context scenarios. Tests on GB300 NVL72 showed single-GPU throughput of more than 16,000 tokens per second and single-user throughput of more than 200 tokens per second. The model also supports inference frameworks including SGLang, vLLM, and TensorRT-LLM.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
