# neuralmagic/deepsparse

Sparsity-aware deep learning inference runtime for CPUs

Repository: https://github.com/neuralmagic/deepsparse
Canonical: https://ross.abutalabs.com/products/deepsparse
Homepage: https://neuralmagic.com/deepsparse/
Language: Python
License: NOASSERTION
License Family: other
Topics: machinelearning, onnx, inference, computer-vision, object-detection, pruning, quantization, pretrained-models, nlp, cpus, sparsification, llm-inference, performance, deepsparse
Archived: true
Last push: 2025-06-02T21:11:06+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 24, release rhythm 8, longevity 100
- inputs: {"age_days": 2088, "days_push": 457, "days_rel": 457, "gap_med": null, "n_releases_24m": 1}
- flags: archived, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3158, forks 193 (observed 2026-08-28T04:07:46.479056+00:00)

## What it is
DeepSparse is a sparsity-aware deep learning inference runtime that delivers GPU-class performance on x86 CPUs for ONNX models, covering computer vision, NLP, and LLM workloads. Development ceased and the project was deprecated on June 2, 2025 following Neural Magic's acquisition by Red Hat.

## Use cases
- run ONNX models fast on CPUs without a GPU
- deploy sparse pruned and quantized models for inference
- serve object detection models on x86 servers
- run LLM inference on CPU hardware
- integrate ML inference into Python applications

## When to choose
- you need fast CPU-only inference for ONNX models and are comfortable with an unmaintained runtime
- you already have sparsified models from Neural Magic's tooling and accept no future updates

## When to avoid
- you need ongoing support, updates, or security patches
- you are starting a new project - use vLLM or other actively maintained runtimes instead
- you need GPU or non-x86 (e.g. ARM) inference

## Facets
- artifact type: library
- maturity: abandoned
- function: llm-inference, machine-learning, computer-vision, nlp
- domain: machine-learning, deep-learning, large-language-models, computer-vision
- platform: python, cpp
- tags: onnx, inference-runtime, sparsity, cpu-optimization, deprecated, pruning, quantization, natural-language-processing, linux

## Member repositories
- neuralmagic/deepsparse (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:46.479056+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:45:33.506841+00:00, confidence not recorded.
  - readme: https://github.com/neuralmagic/deepsparse (fetched 2026-08-28T04:07:46.479056+00:00, sha f0800317b21f)
  - homepage: https://neuralmagic.com/deepsparse/ (fetched 2026-08-29T09:40:05.318525+00:00, sha 8c6922ebda3f)
  - site_page: https://www.redhat.com/en/about/press-releases/red-hat-ai-factory-nvidia-accelerates-path-scalable-production-ai (fetched 2026-08-29T09:40:05.324143+00:00, sha e10bc40cf2db)
  - registry_pypi: https://pypi.org/pypi/deepsparse/json (fetched 2026-08-29T09:40:05.326446+00:00, sha ea18c11fae32)
- Data as of 2026-08-30T08:39:29.467469+00:00.