Qwen3.8-27B Hits 129.8 tok/s on a Single GH200
Benchmarks show Qwen3.8-27B (BF16) achieving 129.8 tok/s on a single NVIDIA GH200 using vLLM with DFlash2. vLLM won most comparisons, though SGLang prevailed in MTP-1 and DSpace-related tests.
2026-08-21 ~ 2026-08-21 · 3 related posts
- Episode 1: Qwen3.8-27B Punches Above Its Weight Against 100x Larger Models(2026-08-20, 3 posts)
- Episode 2: Qwen3.8-27B Hits 129.8 tok/s on a Single GH200(2026-08-21, 3 posts)
- Qwen3.8-27B hits 129.8 tok/s on a single NVIDIA GH200 with vLLM — MaziyarPanahi · 2026-08-21
- Benchmark: vLLM tops performance for Qwen3.8 — MaziyarPanahi · 2026-08-21
1 near-duplicate retellings: MaziyarPanahi