Running DeepSeek V4-Flash Locally: Dual RTX 6000 Rig Handles Only Single User
dee_hw · x · 2026-08-02
A developer shared hands-on experience running DeepSeek V4-Flash on a 2× RTX PRO 6000 build.
- VRAM limits: VRAM caps the context at 131K. This means if two requests each exceed 65K context, one has to wait for the other to finish.
- Context usage: A coding task typically runs >60K context, an agentic task 30–40K, and a chat app 10–15K.
- Verdict: This rig is great for a single personal user and fine for a small team, but it is not suitable as a production box.
Related event: DeepSeek V4 Flash Sparks a Wave of Local Deployment Tests(26 posts)→
More from Infra
- NVIDIA quietly delists the entry-level RTX 5050 from its marketplace, following the RTX 5090 — chemist_slime · 2026-10-16
- Grok claims 816B MoE model runs locally on dual M5 Ultra Mac Studios via MLX — grok · 2026-10-16
- Same 1,000 Cloudflare Durable Objects: $4,050/month cut to $22 with one trick — threepointone · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- Baidu's Virtual Head Attention tames the KV cache, FLOPs, quality triangle in long-context LLMs — baidu · 2026-10-16
- DuoDiscover agents auto-build fast eval proxies, scoring 2.68x TPU kernel speedup — berkeley_ai · 2026-10-16