Enterprise DNAEnterprise DNA

Open Source

MInference

[NeurIPS'24 Spotlight, ICLR'25, ICML'25] To speed up Long-context LLMs' inference, approximate and dynamic sparse calculate the attention, which reduces inference latency by up to

Our verdict

Not reviewed yet

We have not reviewed MInference yet

No one at Enterprise DNA has checked this listing against its sources. The description comes from public listings gathered automatically. Any fact below shows where it came from and when it was last checked.

How we review records

From the listing

About MInference

MInference is a framework that speeds up long-context large language model inference by approximating the attention computation with dynamic sparse patterns. It targets the pre-filling phase and can reduce latency by up to 10x on an A100 GPU while preserving accuracy. The tool is implemented in Python and open source under the Microsoft organization.

Best for

Developers optimizing long-context LLM inference on NVIDIA GPUs

Use cases

  • Accelerating pre-filling for long-context LLM inference pipelines
  • Reducing latency in applications that process large input sequences
  • Evaluating sparse attention as a drop-in replacement for full attention

Notes

1,217 stars on GitHub. Last updated 2026-04-08. Licensed MIT.

Pros

  • Up to 10x latency reduction for pre-filling on A100 hardware
  • Maintains accuracy despite sparse approximation
  • Open source with a published NeurIPS spotlight paper

Cons

  • Optimizations are limited to the pre-filling phase, not generation
  • Requires integration into existing inference codebases
  • Performance gains depend on specific model architectures and hardware

Key facts

What we have checked

Latest version
0.1.6 (released 17 June 2025)
Checked GitHub
Days since the last commit
35
Checked GitHub
GitHub forks
85
Checked GitHub
GitHub stars
1,231
Checked GitHub
Days to a first reply on issues
342.3
Checked GitHub
Open issues
87
Checked GitHub
Releases in the last 90 days
0
Checked GitHub

Each value shows when we last confirmed it and where it came from. Values we have not confirmed are left out, never guessed.

Stars and usage

How MInference is growing

GitHub stars

1,217 to 1,231

stars. 12 July 2026 to 14 Nov 2026, 5 readings.

Source: GitHub. Checked .

Health and security

Is MInference looked after, and what do its setup files show

Health score

56 out of 100

Does someone look after this repository? It does not say the code is good or safe. Worked out from GitHub signals checked . How it is worked out

PartPointsWhat we saw
Recent activity29.6 of 30Last push 35 days ago
Releases0 of 150 releases in 90 days
Issue replies0 of 20First reply in 342.3 days (median)
Maintainer15 of 15Owned by an organisation
Official sourceNot countedMCP servers only

Alternatives

Alternatives to MInference

RecordTypeHow we knowChecked
DeepSpeedThis is an alternative to itOpen sourceNamed in the listingNot dated
SGLangThis is an alternative to itOpen sourceNamed in the listingNot dated
TensorRT-LLMThis is an alternative to itOpen sourceNamed in the listingNot dated
vLLMThis is an alternative to itOpen sourceNamed in the listingNot dated

Compare

Compare MInference side by side

Sources

Where this comes from

  • GitHubRegistry

    Checked · Days since the last commit, GitHub forks, GitHub stars, Days to a first reply on issues, Open issues, Releases in the last 90 days

  • GitHubOfficial

    Checked · Latest version

A source shows what was published when we checked it. It does not certify security, performance or a working installation. Read our method.

Changelog

What changed

  1. Released 0.1.6

    Source: GitHub

  2. Released 0.1.5.post1

    Source: GitHub

  3. Released 0.1.5

    Source: GitHub

  4. Released 0.1.4.post4

    Source: GitHub

  5. Released 0.1.4.post3

    Source: GitHub

  6. Released 0.1.4.post2

    Source: GitHub

  7. Released 0.1.4.post1

    Source: GitHub

  8. Released 0.1.4

    Source: GitHub

  9. Released 0.1.3

    Source: GitHub

  10. Released 0.1.2

    Source: GitHub

Free 27-page guide

Get the free Developer’s Field Guide

A 27-page field guide to the AI coding workflow with Claude. Claude Code, MCP servers, the prompt patterns that work, and what to delegate. Free.

Enter your work email. We send it straight over, plus a few short notes worth knowing. Unsubscribe any time.

No spam. Unsubscribe any time.