Fri, Sep 18 11:14 PM

DeepSeek v4.1 Flash avg 102 tps on 4x RTX6000 pro max-q, 2.1x up from v4-flash

Hacker News•Fri, Sep 18, 2026, 09:27 PM•2 min read

2 points, 2 comments on Hacker News

Source: [Hacker News](https://forum.level1techs.com/t/llm-inference-workstation-4x-rtx6000-blackwell-pro-max-q-384gb-vram-threadripper-pro-7985wx-wrx90e-sage-se-512gb-ram-1700w/252181)

📰 Read Full Story

This is an aggregated headline summary. For the complete report, visit the original publisher.

Continue Reading at Hacker News ↗

#tech #flash #deepseek #avg #tps #rtx #pro #max

More Headlines

TechnologyHacker News• 4m ago

Show HN: CRT – a local code review tool for agentic development

I built crt to help reduce the burden of reviewing AI generated code. As fun as it is to vibecode software to production, sometimes you still need to understand and be responsible for the code that agents write. "It's the agents fault" doesn't work as an excuse for many types of software.

Technology9to5Mac• 5m ago

The Quantified Scientist tests the Apple Watch Series 12’s new heart rate sensor

In a new video, The Quantified Scientist’s Rob ter Horst compares the new Apple Watch Series 12’s heart rate sensor against a Polar H10 ECG chest strap. Here’s what he found.

TechnologyHacker News• 6m ago

macOS Golden Gate – What's New (Roundup)

1 points, 0 comments on Hacker News

TechnologyHacker News• 8m ago

Faster JSON parsing with SVE2 on ARM processors

1 points, 0 comments on Hacker News

TechnologyHacker News• 8m ago

(sixteenth RacketCon) tickets on sale now

1 points, 3 comments on Hacker News

TechnologyHacker News• 9m ago

Enhance Page Elements with Custom Glow Effects and Masks

1 points, 0 comments on Hacker News