Appearance
Models: Foundry Local
Compare this target against the other two
The model availability matrix puts every model on all three targets side by side, and shows the four models that run on all three.
Hardware requirements and sizing maps the model classes on this page to CPU, RAM, GPU, NPU, and storage planning profiles.
Scope
This is the models page for Foundry Local, one of the three targets in ADR-0011. Compare all three targets on the Deployment targets hub.
This target has not been built yet
Nothing on this target has been deployed from this repository, and no automation exists for it yet. Everything below is drawn from accepted decisions and first-party research. Treat it as a design, not an as-built record.
What is decided today
- The Foundry Local catalog only. None of this repository's cloud roster runs here: no image generation, no text to speech, no video.
- The committed first increment is CPU-only, in the
Phi-4-mini-instruct-generic-cpuclass, roughly 4.8 GB, onCPUExecutionProvider. - Microsoft publishes no CPU, RAM, or disk minimum for Foundry Local on Windows at all. The prerequisites name an OS build, a .NET SDK, and a GPU, and nothing else. Any number below is transferred from an adjacent source and labelled as such (SPIKE-25).
- A practical ceiling of roughly a 5 GB quantized 4B-class model was measured on one 64 GB host. Treat that as a single observation, not a rule. SPIKE-18's "core count, not RAM" claim is not supported by any first-party statement, and Microsoft's own troubleshooting guidance points the other way.
- ONNX Runtime supports CPU, CUDA GPU, AMD Vitis NPU, Qualcomm QNN NPU, Intel OpenVINO, WebGPU, and TensorRT RTX execution providers. The NPU and generic-GPU variants are exclusive to this target; Azure Local Foundry never syncs them.
The research and decisions behind this page
These were open when this page was first written. They are now written and linked below. Where a finding contradicted an earlier claim on this page, the page has been corrected and says so inline.