Qwen 3.8 27B served at ~300 tokens/second on Nebius, claims fastest deployment yet
Arindam_1729 · x · 2026-10-14
@1littlecoder reports that Qwen 3.8 27B is currently being served on Nebius at roughly 300 tokens per second, calling it the fastest deployment of the model so far. Hardware details and pricing were not disclosed.
More from Infra
- BitRun runs 1-bit LLMs at 127 t/s on M4 Pro CPU, zero compile, after fixing BitNet kernel bugs — The_VibePreneur · 2026-10-14
- 176K context at 36 t/s on 8GB VRAM: full expert-cache config shared — campaigner_ · 2026-10-14
- Access an NVIDIA DGX Spark from anywhere with NetBird, zero open ports — aestheticedwar1 · 2026-10-14
- Forward Ollama from an NVIDIA DGX Spark to your laptop over SSH — aestheticedwar1 · 2026-10-14
- WireGuard + DGX Spark: give any device secure access to 128GB of local LLM memory — aestheticedwar1 · 2026-10-14
- Qdrant tests Perplexity's pplx-embed-v2-late: 21x smaller vectors, 99.9% retrieval quality kept — qdrant_engine · 2026-10-14