Local benchmark: DeepSeek v4.1 Flash slower than Qwen3.8 Flash-Next at similar quality
Critical-Entry3377 · reddit · 2026-11-14
The author benchmarked DeepSeek v4.1 Flash locally against Qwen 3.8 Flash-Next and found DeepSeek not better for his workload yet much slower, concluding Strata + Qwen 3.8 Flash-Next is the best bang for buck on his machine.
The attached chart quantifies it: DeepSeek-V4.1-Flash scores 63.6 at 44.5 tok/s decode, Qwen3.8-Flash-Next Swift 55.1 at 111.8 tok/s, GLM-5.3-Flash 57 at 30-55 tok/s, and Qwen3.8-27B 50 at 61→141 tok/s (vLLM). DeepSeek is slightly smarter but decodes at 40% of Qwen Flash-Next's speed.
More from Models
- AssemblyAI Adds GPT-6.1 Terra, Beam-1, DeepSeek V4.1 Flash to LLM Gateway — AssemblyAI · 2026-11-14
- Continual Learning Won't Block Full Research Automation — but Will Block Enterprise Adoption — ChrisGPT · 2026-11-14
- Arena weekly: Claude Fable 5.2 tops Agent Arena, Gemini 4 Hydrogen fastest — arena · 2026-11-14
- Open models trail frontier by six months, but that matters less than you think — latkins · 2026-11-14
- Japan's ai& launches Kimi K3.1 Pro and Meta's Muse Spark Open with day-0 pricing — DavidBennett__ · 2026-11-14
- Alan Thompson memo: 18 models in 13 days, Gemini 4 elements, RLI heading to 98% — dralandthompson · 2026-11-14