Local Run of Suspected DeepSeek-4.1-Flash Hits 670 TPS on 4x RTX Pro 6000
Kyrannio · x · 2026-10-23
A user demoed a model labeled DeepSeek-4.1-Flash-1017 running locally on 4x RTX Pro 6000 Max-Q at full precision, hitting 670 TPS on codegen decode using SSD ngram caching, 7-step speculative decoding and a custom codegen-tuned kernel. The model name is unverified by DeepSeek; details may change with an official release.
Related event: DeepSeek-4.1-Flash Hits ~676 tok/s Locally on 4× RTX PRO 6000(2 posts)→
More from Infra
- Coolify: the 63.5k-star open-source self-hosted PaaS that undercuts Vercel on a $5 VPS — thisguyknowsai · 2026-10-23
- ChinaTalk asks: where is China's Groq as inference silicon architectures diverge? — ChinaTalk · 2026-10-23
- Dev open-sources daily updater script that uses LLMs to catch update failures — aronchick · 2026-10-23
- Google Cloud London outage hit banking, delivery and public services, exposing cloud resilience gaps — DavidLinthicum · 2026-10-23
- Samsung Electro-Mechanics joins Nvidia's substrate supply chain, shocking Taiwanese rivals — zephyr_z9 · 2026-10-23
- AI agents drove this dev's GitHub Actions bill from $738 to ~$2,400/month — CellistOk6782 · 2026-10-23