generated: '2026-08-06'
method: searched
source: Docker Hub, Hugging Face, PyPI, npm registry queries + amperecomputing.com/developers
note: 'Ampere Computing publishes no API client libraries — it has no public API. What
  it does publish first-party are runtime artifacts: Ampere Optimized AI framework
  container images and optimized model weights. These are recorded as Packages; NO
  `SDKs` pointer is wired in apis.yml, because none of these is an API client library
  and claiming otherwise would be false credit on the sdk_count check.'
packages:
- language: container
  registry: dockerhub
  namespace: amperecomputingai
  name: amperecomputingai/llama.cpp
  url: https://hub.docker.com/r/amperecomputingai/llama.cpp
  install: docker pull amperecomputingai/llama.cpp
  official: true
  pulls: 15103
  description: Ampere Optimized llama.cpp for arm64 Ampere processors
- language: container
  registry: dockerhub
  namespace: amperecomputingai
  name: amperecomputingai/pytorch
  url: https://hub.docker.com/r/amperecomputingai/pytorch
  install: docker pull amperecomputingai/pytorch
  official: true
  pulls: 6112
  description: Ampere Optimized PyTorch
- language: container
  registry: dockerhub
  namespace: amperecomputingai
  name: amperecomputingai/tensorflow
  url: https://hub.docker.com/r/amperecomputingai/tensorflow
  install: docker pull amperecomputingai/tensorflow
  official: true
  pulls: 4629
  description: Ampere Optimized TensorFlow
- language: container
  registry: dockerhub
  namespace: amperecomputingai
  name: amperecomputingai/onnxruntime
  url: https://hub.docker.com/r/amperecomputingai/onnxruntime
  install: docker pull amperecomputingai/onnxruntime
  official: true
  pulls: 2655
  description: Ampere Optimized ONNX Runtime
- language: container
  registry: dockerhub
  namespace: amperecomputingai
  name: amperecomputingai/ollama
  url: https://hub.docker.com/r/amperecomputingai/ollama
  install: docker pull amperecomputingai/ollama
  official: true
  pulls: 810
  description: Ampere Optimized Ollama
model_artifacts:
- registry: huggingface
  org: AmpereComputing
  url: https://huggingface.co/AmpereComputing
  official: true
  model_count: 68
  api: https://huggingface.co/api/models?author=AmpereComputing
  description: 'Ampere-published GGUF model weights tuned for Ampere CPUs (e.g.
    llama-3.3-70b-instruct-gguf, llama-3.1-8b-gguf, mistral-v0.3-7b-instruct-gguf).'
source_code:
- registry: github
  org: AmpereComputing
  url: https://github.com/AmpereComputing
  official: true
  repo_count: 88
  note: 'Kernel/QEMU/EDK2 staging trees, Terraform modules for OCI/GCP Ampere instances,
    ampere-performance-toolkit, ampere-porting-advisor, ampere-pmu-profiler,
    ampere-gcc / ampere-glibc / ampere-binutils. No OpenAPI, Swagger, AsyncAPI or
    apis.json file is present anywhere in the org (GitHub code search, 0 hits).'
negative_findings:
- registry: pypi
  probed: ampere-computing, ampere-porting-advisor
  result: 404 / not first-party
  note: 'The PyPI package `ampere` exists (200) but is an unrelated battery-modelling
    library by Neal Dawson-Elli — explicitly NOT Ampere Computing.'
- registry: npm
  probed: '@amperecomputing/sdk, ampere-computing'
  result: 404