Red Hat AI(@RedHat_AI):Now stack it with quantization. Pair the speculator above with our NVFP4 checkpoint: a 4-bit, Blackwell-native target, with DSpark's faster decoding on top. Serve RedHatAI/Qwen3.8-27B-NVFP4 with the DSpark speculative-config (method dspark, 8 tokens).

Red Hat AI

@RedHat_AI

Accelerating AI innovation with open platforms and community. The future of AI is open.

加入 May 2018

2.1K 正在關注 12.6K 粉絲

Red Hat AI@RedHat_AI

2026.09.24 16:30

Now stack it with quantization. Pair the speculator above with our NVFP4 checkpoint: a 4-bit, Blackwell-native target, with DSpark's faster decoding on top. Serve RedHatAI/Qwen3.8-27B-NVFP4 with the DSpark speculative-config (method dspark, 8 tokens).

Red Hat AI@RedHat_AI

2026.09.23 15:46

🚀 Our DSpark for Qwen3.8-27B beats native MTP with the same 8 speculative tokens on 4×H200. Up to 52% faster single-stream decoding and 23% higher peak throughput. On 8-needle MRCR, it averages 4.30 accepted tokens beyond 1M context. Model: Demo below, see for yourself!