---
format: "aidr-story-markdown/v1"
id: "d67ea427b92ebf5bc26f399e11e98cbdd42eb270aaa933870a7129b35193d3f6"
canonical_url: "https://aidr.today/d67ea427?lang=en"
title: "New Rust-Based Inference Engine for DeepSeek v4.1-Flash Reaches 2060 TPS"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-13T21:07:57.000Z"
category: "Open Source"
topics: ["deepseek","inference","open-source"]
source_urls: ["https://marketbrief.now/ai/new-rust-based-inference-engine-for-deepseek-v41-flash-reaches-2060-tps-7ae86031","https://huggingnews.com/ai/new-rust-based-inference-engine-for-deepseek-v41-flash-reaches-2060-tps-7ae86031"]
summary: "A custom inference tool for the DeepSeek v4.1-flash model was built using Rust and CUDA. This engine reaches a decode speed of about 2060 tokens per second on 8 B300 GPUs. The project also records a 1M token prefill in approximately 12 seconds on 4 B300 GPUs."
---

# New Rust\-Based Inference Engine for DeepSeek v4\.1\-Flash Reaches 2060 TPS

> [Open the canonical story](<https://aidr.today/d67ea427?lang=en>)

**Published:** 2026-11-13T21:07:57.000Z
**Category:** Open Source
**Topics:** deepseek, inference, open\-source

## Summary

A custom inference tool for the DeepSeek v4\.1\-flash model was built using Rust and CUDA\. This engine reaches a decode speed of about 2060 tokens per second on 8 B300 GPUs\. The project also records a 1M token prefill in approximately 12 seconds on 4 B300 GPUs\.

## Sources

- [Story source](<https://marketbrief.now/ai/new-rust-based-inference-engine-for-deepseek-v41-flash-reaches-2060-tps-7ae86031>)
- [Story source](<https://huggingnews.com/ai/new-rust-based-inference-engine-for-deepseek-v41-flash-reaches-2060-tps-7ae86031>)