FULL STORY
Qwen3.8-27B: Small Model Wows in Early Tests
Alibaba's Qwen3.8-27B drew praise after release, with users running Q3 quantized builds on consumer GPUs for serious coding tasks, and benchmarks showing 129.8 tok/s on a single GH200.
2026-08-20 ~ 2026-08-22 · 2 episodes · 6 posts
Episode 1 · Qwen3.8-27B Punches Above Its Weight Against 100x Larger Models (2026-08-20, 3 posts)
Qwen3.8-27B is drawing praise for its real-world performance: users ran the Q3 quantized version on an RTX 4060 Ti for serious coding tasks, while Arena's first tests compared it against models 100x its size. Developers say it is optimized to actually get work done.
- Dev praises Qwen 3.8 27B: Benchmarked to actually work — rudrank · 2026-08-20
- Qwen 3.8 27B Q3 Review: Strong Performance Despite Low Quantization — AltruisticList6000 · 2026-08-21
- Qwen3.8-27B tested against models 100x its size in first impressions video — arena · 2026-08-22
Episode 2 · Qwen3.8-27B Hits 129.8 tok/s on a Single GH200 (2026-08-21, 3 posts)
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.
- Qwen3.8-27B hits 129.8 tok/s on a single NVIDIA GH200 with vLLM — MaziyarPanahi · 2026-08-21
- Qwen3.8-27B hits 129.8 tok/s on one GH200; vLLM wins — MaziyarPanahi · 2026-08-21
- Benchmark: vLLM tops performance for Qwen3.8 — MaziyarPanahi · 2026-08-21