Early Qwen3 27B dense model benchmark: 671 tok/s prefill, 22 tok/s decode, 2.4s first token
draginol · x · 2026-10-18
Blogger draginol shared early benchmark numbers for the dense Qwen 3.8 27B model: 671 tok/s prefill at 1K context (with some skipped), 22 tok/s decode streaming speed, and a 2,393 ms first-token latency. He also linked a workaround tweak for an issue encountered during testing, and says more extensive benchmarking is coming soon.
More from Infra
- Developer builds change alerts for LLM prices and retirements across Bedrock, Azure and Vertex — jsandhu0780 · 2026-10-18
- 12.3M tokens in 27 minutes for $0.14: agent session with 99.6% cache hit — teortaxesTex · 2026-10-18
- Running Qwen3 27B at 262k context on a single 16GB RX 9060 XT at 42.7 tps — Maglcite · 2026-10-18
- SlateDB+Lance+Datafusion stack saves customer millions, as AI agents unbundle big data systems — sh_reya · 2026-10-18
- Paying $0.50 per 10 seconds for video gen APIs — when is self-hosting worth it? — Key-Lychee-913 · 2026-10-18
- vLLM and NVIDIA push Kimi K3 decode to 842 tokens/s on GB300 with fused CuTe DSL path — vllm_project · 2026-10-18