Thursday, 22 October 2026
Anawords
Research desk
Research noteResearched 22 Oct 2026 · published 22 Oct 2026, 20:09 IST

Research note: DeepSeek open-sources DeepInfer, the engine that serves its V4.1 Flash model

How we checked DeepInfer’s release, hardware support and what was left out.

Research led by Mayank Sahu
Compiled with AI assistance (Claude by Anthropic)
Read the story →

At a glance

1searches run
3sources read
2sources used
6facts labelled
Strongevidence strength · 4 of 5 facts (80%)

How we rate evidence: Most facts come from official sources or two or more independent outlets. We count the share of known facts labelled Confirmed or Reported: 70% or more is Strong, 40–69% Mixed, below 40% Limited.

From sources to story

How many sources we read, how many we used, and how many are official or archived.

Read3Every source we opened
Used2Relied on for the story
Cited2Linked in the story
Official0From the company or a public body
Archived2Saved on the Internet Archive

How sure we are, fact by fact

  • Reported · 4 — Two or more independent, named outlets say it.
  • Company claim · 1 — Only the company says it. Not checked by anyone else yet.
  • Not known · 1 — Not announced or not answered yet.

Evidence map

Each row is a fact from the story; each column is a source. A filled dot means that source supports the fact.

FactLabel1OUT2OUT
DeepSeek open-sourced DeepInfer under Apache 2.0 on 20 October 2026.Reported
It supports only DeepSeek-V4.1-Flash for now.Reported
It runs on Nvidia SM100 Blackwell GPUs and Huawei Ascend 950 chips.Reported
Cache management and load-balancing code were not released.Reported
DeepSeek reports a 4% to 7% throughput gain at production load.Company claim
Independent performance results.Not known
Question

What we set out to check

What DeepInfer includes, its licence, hardware support and what was held back.

Findings

What we found

  • Both outlets agree on the Apache 2.0 licence, single-model support, hardware and the excluded components.
  • SiliconANGLE gives DeepSeek’s performance and revenue figures.
Discrepancies

Numbers that differ

None.

Method

Labels we chose and why

  • Reported for release details in both outlets.
  • Claim for DeepSeek’s performance figures.
Open questions

What we could not verify

  • The GitHub README directly.

Can you help confirm one of these? Send us a source. When we find an answer, we update this note and the story, with the date.

Next steps

How we could improve this

Read the repository README and licence file directly, and add official confirmation of supported hardware. We would also look for independent benchmarks from the vLLM or SGLang communities, and check whether DeepSeek publishes a roadmap for the components it kept private, such as cache management and load balancing.

Research timeline

  1. Research22 Oct 2026
  2. Research note published22 Oct 2026, 20:09 IST
  3. Story published22 Oct 2026, 20:19 IST

What we searched

  • DeepSeek DeepInfer open source inference framework

Every source we read

SourceUsed?Why
SiliconANGLE
Archived copy		USEDComponents, hardware, figures.
TNW
Archived copy		USEDOrigins, exclusions, context.
DeepSeek on XNOT USEDCould not open.

Sources cited in the story

  1. 1SiliconANGLE: DeepSeek open-sources DeepInfer, the inference framework behind its production servicesOUTLET
  2. 2TNW: DeepSeek open-sources DeepInfer, the engine that runs its AI modelsOUTLET

Research notes show our working. They are not peer-reviewed papers. Data: facts as JSON · How we report