| PC: CPU only, WSL | 1.7B INT8 historical M0 pass; extraction rerun pending | Not qualified | Not qualified | Not qualified | Not qualified |
| PC: CPU only, native Windows | Not qualified | Not qualified | Not qualified | Not qualified | Not qualified |
| PC: CPU + NVIDIA discrete GPU, Windows/WSL | 4B BF16 M0 passed | Real-weight streaming completes; playback stalls remain | Complete multimodal pipeline not qualified | Not qualified | Not qualified |
| PC: CPU + AMD Radeon 890M iGPU | Not qualified | Not qualified | Historical vision-tower component pass only | Not qualified | Not qualified |
| PC: CPU + AMD NPU | Not qualified | Not qualified | Tested whole-vision-tower exports rejected | Not qualified | Not qualified |
| PC: CPU + iGPU + NPU, with optional discrete GPU used jointly | Full heterogeneous model path not qualified | Full heterogeneous model path not qualified | Full heterogeneous model path not qualified | Not qualified | Not qualified |
| Mobile: Galaxy S25 / Snapdragon 8 Elite for Galaxy, CPU + GPU + NPU | W8A16 attention component passed in AI Hub; full local generation pending | Local NPU/GPU pipeline not qualified | Not qualified | Not qualified | Not qualified |
| Other mobile / embedded SoCs | Not qualified | Not qualified | Not qualified | Not qualified | Not qualified |
The PC measurements below come from one Ryzen AI 9 HX370 laptop. Component
success, backend probes and topology tests do not certify a full model pipeline.
S25 AI Hub results do not qualify an on-device Omni controller, continuous
generation, shared NPU/GPU execution or sustained thermal behavior.
The tested PC is a Ryzen AI 9 HX370 laptop with Radeon 890M, AMD NPU and
RTX 5090 Laptop GPU. Selecting CPU-only or a subset of those accelerators is
an execution configuration on that PC, not evidence from a different PC SKU.
Model support is specific to the following pairings:
| Model / artifact | PC execution configuration | OS / backend | Qualification |
|---|---|---|---|
| Spark-X2.5-4B BF16 | CPU + RTX 5090 Laptop GPU | WSL / vLLM CUDA | Current branch: M0 text/state/cancellation workload passed. |
| Spark-X2.5-4B BF16 | CPU + RTX 5090 Laptop GPU | Native Windows / vLLM CUDA | Current branch: M0 text/state/cancellation workload passed. |
| Spark-X2.5-1.7B INT8 | CPU execution; accelerators unused | WSL / vLLM CPU | Historical M0 acceptance and token parity passed on 2026-09-15; not freshly requalified on this extraction branch. |
| Spark-X2.5-1.7B INT8 | CPU execution; no NVIDIA dependency | Native Windows | Model path not yet qualified; CPU Add graph success is not text-model support. |
| Qwen3-TTS-12Hz-0.6B CustomVoice | CPU + RTX 5090 Laptop GPU; both stages on GPU | WSL / vLLM CUDA | Current branch: real-weight audio streaming completes; playback stalls and remaining M2 gates stay open. |
| Qwen3-TTS-12Hz-0.6B CustomVoice | CPU + RTX 5090 Laptop GPU; both stages on GPU | Native Windows / vLLM CUDA | Current branch: real-weight audio streaming completes; playback stalls and remaining M2 gates stay open. |
| Qwen3.8-27B vision tower, FP32 ONNX | CPU + Radeon 890M iGPU | Windows / ORT DirectML | Historical component validation only; not the complete 27B model or a newly qualified graph-stage vision-to-language pipeline. |
| Qwen3.8-27B vision tower, tested A16W8 exports | CPU + AMD NPU | Windows / ORT VitisAI | Tested whole-tower artifacts were rejected by NPU partitioning. This does not rule out other exports or smaller components. |
| Qwen3.8-27B complete multimodal pipeline | CPU/iGPU/NPU, with optional discrete GPU | Target-specific | Not qualified by this abstraction release; earlier short-text loading is not complete multimodal validation. |
| MiniCPM-o | CPU/iGPU/NPU, with optional discrete GPU | Target-specific | Full model × PC combinations remain unqualified for this release. |
| InternVLA | CPU/iGPU/NPU, with optional discrete GPU | Target-specific | Full model × PC combinations remain unqualified for this release. |
ORT Add and the small AMD NPU graph are backend probes, so they are deliberately
not listed as supported application models. Likewise, a Galaxy S25 Spark attention
component run in AI Hub is a separate **model component × phone** pairing, not a
PC pairing or proof of complete mobile Spark execution. Other model × PC × backend
combinations require their own artifact and workload qualification.
### Hardware layouts represented by the abstraction
The abstraction supports these single-machine compositions. This table separates
topology/admission support from successful execution of a particular model; it
does not promise that every listed accelerator can run every model.
| Composition | Resource accounting | Current execution evidence |
|---|---|---|
| CPU only | Host RAM; WSL quota when applicable | ORT CPU graphs through public Omni on Windows and WSL. Native Windows CPU text remains unqualified. |
| CPU + integrated GPU | Shared host RAM | Radeon 890M DirectML graph through Omni. This extraction's probe is not a complete vision/language pipeline. |
| CPU + integrated NPU | Shared host RAM | AMD NPU VitisAI graph through Omni; permitted CPU partitions are explicit in the placement report. |
| CPU + integrated GPU + integrated NPU | Shared RAM plus optional observed package/power/bandwidth relationships | Topology and admission contracts tested; individual laptop routes execute. Simultaneous full-model overlap is not qualified. |
| CPU + integrated GPU + integrated NPU + discrete GPU | Shared RAM and separate discrete VRAM | Current Ryzen AI 9 HX370 / Radeon 890M / AMD NPU / RTX 5090 Laptop machine. Spark CUDA text and native two-stage TTS run on Windows and WSL; using all accelerators together is not qualified. |
| CPU + discrete GPU, with integrated devices unused or absent | Host RAM plus discrete VRAM | Native CUDA text/TTS execution uses this subset. NVIDIA is optional in the abstraction. |
Mobile/embedded SoCs fit the CPU + integrated GPU + NPU composition, but their
local controller and model paths still require device-specific qualification.
Galaxy S25 AI Hub inference/profiling passed for one Spark attention component;
this is component evidence, not full local mobile generation or streaming.
AI Hub is a test facility, never a deployment dependency. Intel, Apple and other
vendor combinations are not certified by the AMD/NVIDIA/Qualcomm results.
`DeviceDescriptor` separates physical device identity, execution domain and
integration. A CPU, integrated GPU and NPU can all reference `machine:ram`.
An additional discrete GPU references its own `machine:vram:<physical-id>`.
A WSL allocation can consume both `machine:ram` and `wsl:quota` constraints;
those are two limits on the same allocation, not two amounts of physical RAM.
Use the same physical GPU ID when Windows and WSL expose the same adapter.
Package IDs, bandwidth groups and power domains are optional observed relationships.
Missing values mean unknown. Fitting memory does not authorize overlap on a shared
package: selection and overlap still require model quality and complete workload
measurements. Synthetic topology tests cover CPU-only and all four integrated /
discrete combinations; they are not hardware certification.
## Deploy a graph
Register a `PipelineConfig` with `StageExecutionType.GRAPH`, `model_stage="graph"`
and an appropriate final output type. Set `async_chunk=False`. A graph following
another stage must have a model-owned `custom_process_input_func` that returns
`{"tensors": {name: numpy_array}}`. Names, dtypes and shapes must match the
manifest's placement examples. Graph stages do not tokenize prompts or apply
model-specific transformations.
Graph-containing v1 pipelines must be a linear chain with exactly one terminal
output at the last stage. Fan-out, joins and multiple final outputs are refused
before loading; those require reference-counted consumer leases.
Each stage in a graph-containing plan declares the same capacities and its own