DeepSeek V4.1 Flash clocks ~1038 tok/s live via Baseten inference
gajesh · x · 2026-11-01
A screenshot shared by gajesh shows DeepSeek V4.1 Flash 1017 Fast, served via Baseten, generating at roughly 1038 tokens per second during a live coding session — a striking throughput figure for a frontier-class open model.
More from Infra
- GPU break-even token throughput up 25-83% since May, squeezing inference providers' margins — AccBalanced · 2026-11-01
- Musk Reveals Macrohard Scale: 110K GB200 and 220K GB300 GPUs — TinfoilTricorn · 2026-11-01
- RTX 4070 runs MiniMax H3: 12s video takes 1.5h, 15s takes 5h — Serenaded · 2026-11-01
- Intel crashes 30% in worst day in 50 years as Anthropic junk-rating rumor spreads — teortaxesTex · 2026-11-01
- 64GB and 96GB RTX 5090 variants confirmed, launching Nov 2026 — lxfater · 2026-11-01
- Sharing a laptop GPU to a desktop, he used local Qwen to build a playable Galaga-like game — draginol · 2026-11-01