Mobile / PC × model qualification audit
2026-09-22: 60 named pairings checked, not 60 successful executions.
PASS = specified full workload passed; PARTIAL = functional output with acceptance gaps; COMPONENT = one model component only; PROFILE_ONLY = component timing without quality; FAILED = attempted execution/startup failed; REJECTED = specific artifact refused; BLOCKED = a named prerequisite is missing. Historical evidence is marked in each finding.
| Device / execution configuration | Spark-X2.5 | Qwen3-TTS 0.6B CustomVoice | Qwen3.8-27B | MiniCPM-o 4.5 | InternVLA-A1 |
|---|---|---|---|---|---|
| PC HX370: CPU / WSL | PASS | FAILED | BLOCKED | BLOCKED | BLOCKED |
| PC HX370: CPU / native Windows | BLOCKED | BLOCKED | BLOCKED | BLOCKED | BLOCKED |
| PC HX370 + RTX 5090 Laptop / WSL | PASS | PARTIAL | FAILED | BLOCKED | BLOCKED |
| PC HX370 + RTX 5090 Laptop / Windows | PASS | PARTIAL | FAILED | BLOCKED | BLOCKED |
| PC HX370 + Radeon 890M / Windows worker | BLOCKED | BLOCKED | COMPONENT | BLOCKED | BLOCKED |
| PC HX370 + AMD NPU / Windows worker | BLOCKED | BLOCKED | REJECTED | BLOCKED | BLOCKED |
| PC CPU+iGPU+NPU +/- discrete GPU, joint model execution | BLOCKED | BLOCKED | BLOCKED | BLOCKED | BLOCKED |
| PC Snapdragon X Elite CRD / AI Hub | BLOCKED | PROFILE_ONLY | BLOCKED | BLOCKED | BLOCKED |
| Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy | COMPONENT | COMPONENT | BLOCKED | COMPONENT | BLOCKED |
| Mobile Galaxy S24 / Snapdragon 8 Gen 3 | BLOCKED | COMPONENT | BLOCKED | COMPONENT | BLOCKED |
| Embedded SA8775P ADP / AI Hub | BLOCKED | PROFILE_ONLY | BLOCKED | BLOCKED | BLOCKED |
| Embedded RB3 Gen 2 / QCS6490 / AI Hub | BLOCKED | PROFILE_ONLY | BLOCKED | BLOCKED | BLOCKED |
Scope and reproduction
PC rows are execution configurations on the HX370/890M/AMD NPU/RTX 5090 Laptop machine. X Elite is a separate Windows ARM device accessed through AI Hub. The two embedded kits are not phones. Other SKUs are outside this finite audit, not implied supported.
The checked roots contain MiniCPM5-2B, which must not substitute for MiniCPM-o 4.5. Missing complete MiniCPM-o and InternVLA artifacts block those local runs. ADB reports no attached phone; Hub component jobs cannot prove a device-local controller or sustained co-residency.
The CPU venv is vLLM 0.28; the CUDA/native-Windows venvs are 0.29. The CPU TTS tuple failure is specific to the mixed-version environment. Preserve failures instead of describing these as universal model limitations.
Raw selected reports are in evidence/. matrix.json retains every source path/hash, cell finding and next step. Fresh probe scripts are in repro/; see their --help and the workspace experiment README for exact environments. No model weights or API credentials are included.
Per-pair results
PC HX370: CPU / WSL × Spark-X2.5
PASS — P. Fresh Spark 1.7B INT8 acceptance: 12 prompts x 128 tokens; clean cancellation. Current Omni branch with the existing vLLM 0.28 CPU wheel; version mismatch warning retained, not a claim about a 0.29 CPU wheel.
Evidence: spark-cpu, spark-cpu-log.
Next: Qualify a matching vLLM 0.29 CPU build and token parity separately.
PC HX370: CPU / WSL × Qwen3-TTS 0.6B CustomVoice
FAILED — execution. Fresh real-weight two-frame probe loaded both stages but generation failed in sanitize_min_tokens_stop_ids: expected four tuple fields, got three. Existing CPU wheel is vLLM 0.28 versus Omni 0.29. No audio pass.
Evidence: tts-cpu, tts-cpu-log.
Next: Use a matching CPU runtime or explicitly validated compatibility fix; repeat real audio, then full streaming gates.
PC HX370: CPU / WSL × Qwen3.8-27B
BLOCKED — none. No compatible CPU artifact/deployment for this model was found in the checked model roots. Available Qwen 27B FP8/NVFP4 CUDA artifacts do not establish a CPU route.
Evidence: host-wsl.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370: CPU / WSL × MiniCPM-o 4.5
BLOCKED — none. No complete matching checkpoint found in the inventoried workspace, HF cache and native model roots. MiniCPM5-2B is not MiniCPM-o 4.5. Runtime registration alone is not a model run.
Evidence: host-wsl, host-windows.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370: CPU / WSL × InternVLA-A1
BLOCKED — none. No complete matching checkpoint found in the inventoried workspace, HF cache and native model roots. MiniCPM5-2B is not MiniCPM-o 4.5. Runtime registration alone is not a model run.
Evidence: host-wsl, host-windows.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370: CPU / native Windows × Spark-X2.5
BLOCKED — none. The installed native Windows vLLM extension lacks cpu_attn_reshape_and_cache and cpu_attn_get_scheduler_metadata. The separate ORT CPU graph worker is not a whole-model text/TTS adapter; a qualified CPU build/binding is missing.
Evidence: host-windows, windows-cpu-kernels, backend-factory.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370: CPU / native Windows × Qwen3-TTS 0.6B CustomVoice
BLOCKED — none. The installed native Windows vLLM extension lacks cpu_attn_reshape_and_cache and cpu_attn_get_scheduler_metadata. The separate ORT CPU graph worker is not a whole-model text/TTS adapter; a qualified CPU build/binding is missing.
Evidence: host-windows, windows-cpu-kernels, backend-factory.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370: CPU / native Windows × Qwen3.8-27B
BLOCKED — none. The installed native Windows vLLM extension lacks cpu_attn_reshape_and_cache and cpu_attn_get_scheduler_metadata. The separate ORT CPU graph worker is not a whole-model text/TTS adapter; a qualified CPU build/binding is missing.
Evidence: host-windows, windows-cpu-kernels, backend-factory.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370: CPU / native Windows × MiniCPM-o 4.5
BLOCKED — none. No complete matching checkpoint found in the inventoried workspace, HF cache and native model roots. MiniCPM5-2B is not MiniCPM-o 4.5. Runtime registration alone is not a model run.
Evidence: host-wsl, host-windows.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370: CPU / native Windows × InternVLA-A1
BLOCKED — none. No complete matching checkpoint found in the inventoried workspace, HF cache and native model roots. MiniCPM5-2B is not MiniCPM-o 4.5. Runtime registration alone is not a model run.
Evidence: host-wsl, host-windows.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + RTX 5090 Laptop / WSL × Spark-X2.5
PASS — P. Spark 4B BF16: 11 M0 tests passed on this branch for text, state, cancellation and memory; scoped to the recorded workload.
Evidence: spark-cuda-wsl.
Next: Broaden only with a named workload/device/precision qualification.
PC HX370 + RTX 5090 Laptop / WSL × Qwen3-TTS 0.6B CustomVoice
PARTIAL — P-limited. Real-weight two-stage audio completes. One measured utterance has playback stalls; speech quality, long streaming, interruption/recovery and thermal gates remain open.
Evidence: tts-cuda-wsl.
Next: Resolve and measure playback stalls; run quality, interruption and thermal tests.
PC HX370 + RTX 5090 Laptop / WSL × Qwen3.8-27B
FAILED — startup. Fresh public Omni startup rejects the real NVFP4 checkpoint: no registered Omni pipeline. Historical bare-vLLM short-text success does not qualify this integration or multimodal inference.
Evidence: qwen-wsl, qwen-text-historical.
Next: Add/validate an Omni pipeline binding; rerun text, image, context and artifact-quality tests.
PC HX370 + RTX 5090 Laptop / WSL × MiniCPM-o 4.5
BLOCKED — none. No complete matching checkpoint found in the inventoried workspace, HF cache and native model roots. MiniCPM5-2B is not MiniCPM-o 4.5. Runtime registration alone is not a model run.
Evidence: host-wsl, host-windows.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + RTX 5090 Laptop / WSL × InternVLA-A1
BLOCKED — none. No complete matching checkpoint found in the inventoried workspace, HF cache and native model roots. MiniCPM5-2B is not MiniCPM-o 4.5. Runtime registration alone is not a model run.
Evidence: host-wsl, host-windows.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + RTX 5090 Laptop / Windows × Spark-X2.5
PASS — P. Spark 4B BF16: 11 M0 tests passed on this branch for text, state, cancellation and memory; scoped to the recorded workload.
Evidence: spark-cuda-windows.
Next: Broaden only with a named workload/device/precision qualification.
PC HX370 + RTX 5090 Laptop / Windows × Qwen3-TTS 0.6B CustomVoice
PARTIAL — P-limited. Real-weight two-stage audio completes. One measured utterance has playback stalls; speech quality, long streaming, interruption/recovery and thermal gates remain open.
Evidence: tts-cuda-windows.
Next: Resolve and measure playback stalls; run quality, interruption and thermal tests.
PC HX370 + RTX 5090 Laptop / Windows × Qwen3.8-27B
FAILED — startup. Fresh public Omni startup rejects the real NVFP4 checkpoint: no registered Omni pipeline. Historical bare-vLLM short-text success does not qualify this integration or multimodal inference.
Evidence: qwen-windows, qwen-text-historical.
Next: Add/validate an Omni pipeline binding; rerun text, image, context and artifact-quality tests.
PC HX370 + RTX 5090 Laptop / Windows × MiniCPM-o 4.5
BLOCKED — none. No complete matching checkpoint found in the inventoried workspace, HF cache and native model roots. MiniCPM5-2B is not MiniCPM-o 4.5. Runtime registration alone is not a model run.
Evidence: host-wsl, host-windows.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + RTX 5090 Laptop / Windows × InternVLA-A1
BLOCKED — none. No complete matching checkpoint found in the inventoried workspace, HF cache and native model roots. MiniCPM5-2B is not MiniCPM-o 4.5. Runtime registration alone is not a model run.
Evidence: host-wsl, host-windows.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + Radeon 890M / Windows worker × Spark-X2.5
BLOCKED — none. No placement- and quality-verified whole-model graph bundle/stateful adapter for this model on this AMD route. Small backend graphs do not establish model support.
Evidence: backend-dml, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + Radeon 890M / Windows worker × Qwen3-TTS 0.6B CustomVoice
BLOCKED — none. No placement- and quality-verified whole-model graph bundle/stateful adapter for this model on this AMD route. Small backend graphs do not establish model support.
Evidence: backend-dml, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + Radeon 890M / Windows worker × Qwen3.8-27B
COMPONENT — C-historical. FP32 448x448 vision tower: DirectML 85/113 nodes; normalized max error 9.095e-06 versus torch CPU. Historical component path, not full 27B or current graph-stage image-to-text integration.
Evidence: qwen-vision-dml.
Next: Recover the hashed tower bundle and qualify its current Omni graph handoff with the language stage.
PC HX370 + Radeon 890M / Windows worker × MiniCPM-o 4.5
BLOCKED — none. No placement- and quality-verified whole-model graph bundle/stateful adapter for this model on this AMD route. Small backend graphs do not establish model support.
Evidence: backend-dml, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + Radeon 890M / Windows worker × InternVLA-A1
BLOCKED — none. No placement- and quality-verified whole-model graph bundle/stateful adapter for this model on this AMD route. Small backend graphs do not establish model support.
Evidence: backend-dml, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + AMD NPU / Windows worker × Spark-X2.5
BLOCKED — none. No placement- and quality-verified whole-model graph bundle/stateful adapter for this model on this AMD route. Small backend graphs do not establish model support.
Evidence: backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + AMD NPU / Windows worker × Qwen3-TTS 0.6B CustomVoice
BLOCKED — none. No placement- and quality-verified whole-model graph bundle/stateful adapter for this model on this AMD route. Small backend graphs do not establish model support.
Evidence: backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + AMD NPU / Windows worker × Qwen3.8-27B
REJECTED — artifact. The tested A16W8 whole-vision-tower export failed NPU placement. This conclusion is specific to that export, not all future Qwen/NPU combinations.
Evidence: qwen-vision-npu.
Next: Validate a supported export or smaller useful component, including same-checkpoint numerical quality.
PC HX370 + AMD NPU / Windows worker × MiniCPM-o 4.5
BLOCKED — none. No placement- and quality-verified whole-model graph bundle/stateful adapter for this model on this AMD route. Small backend graphs do not establish model support.
Evidence: backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC HX370 + AMD NPU / Windows worker × InternVLA-A1
BLOCKED — none. No placement- and quality-verified whole-model graph bundle/stateful adapter for this model on this AMD route. Small backend graphs do not establish model support.
Evidence: backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC CPU+iGPU+NPU +/- discrete GPU, joint model execution × Spark-X2.5
BLOCKED — none. No qualifying same-model multi-accelerator deployment/artifact plan or overlap/recovery evidence. Individual routes are not a joint pipeline.
Evidence: backend-dml, backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC CPU+iGPU+NPU +/- discrete GPU, joint model execution × Qwen3-TTS 0.6B CustomVoice
BLOCKED — none. No qualifying same-model multi-accelerator deployment/artifact plan or overlap/recovery evidence. Individual routes are not a joint pipeline.
Evidence: backend-dml, backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC CPU+iGPU+NPU +/- discrete GPU, joint model execution × Qwen3.8-27B
BLOCKED — none. No qualifying same-model multi-accelerator deployment/artifact plan or overlap/recovery evidence. Individual routes are not a joint pipeline.
Evidence: backend-dml, backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC CPU+iGPU+NPU +/- discrete GPU, joint model execution × MiniCPM-o 4.5
BLOCKED — none. No qualifying same-model multi-accelerator deployment/artifact plan or overlap/recovery evidence. Individual routes are not a joint pipeline.
Evidence: backend-dml, backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC CPU+iGPU+NPU +/- discrete GPU, joint model execution × InternVLA-A1
BLOCKED — none. No qualifying same-model multi-accelerator deployment/artifact plan or overlap/recovery evidence. Individual routes are not a joint pipeline.
Evidence: backend-dml, backend-npu, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC Snapdragon X Elite CRD / AI Hub × Spark-X2.5
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC Snapdragon X Elite CRD / AI Hub × Qwen3-TTS 0.6B CustomVoice
PROFILE_ONLY — C-timing-only. Historical QNN talker profile re-fetched successfully; no numerical/full-pipeline gate. Historical TFLite vocoder target was refused on this Windows ARM device.
Evidence: tts-xelite, tts-xelite-vocoder-rejected, tts-other-devices, tts-other-device-code.
Next: Validate target-specific numerical quality, complete local execution, memory and sustained behavior.
PC Snapdragon X Elite CRD / AI Hub × Qwen3.8-27B
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC Snapdragon X Elite CRD / AI Hub × MiniCPM-o 4.5
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
PC Snapdragon X Elite CRD / AI Hub × InternVLA-A1
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy × Spark-X2.5
COMPONENT — C. Fresh 2026-09-22 W8A16 full-attention component inference/profile passed; full1024 denotes one cache bucket. No full prefill/decode/KV/ring/sampling loop.
Evidence: spark-s25, spark-s25-parity.
Next: Implement and validate local 128-token generation, 512/1024 transitions, cancellation and resident memory.
Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy × Qwen3-TTS 0.6B CustomVoice
COMPONENT — C. Fresh FP16 predictor inference passed; its one output exactly matches the same compiled artifact/dataset reference. Historical decoder parity exists. No complete co-resident NPU/GPU stream.
Evidence: tts-s25, hub-repeatability, tts-mobile-parity.
Next: Validate the device-local talker/predictor/vocoder pipeline, output quality, recovery and 30-minute thermal behavior.
Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy × Qwen3.8-27B
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy × MiniCPM-o 4.5
COMPONENT — C. Fresh FP16 speech-head inference passed; all 41 outputs exactly match the same artifact/dataset reference. Historical FP16 numerical parity passed; tested W8A16 top-1 parity failed. Thinker/encoders/vocoder are not qualified together.
Evidence: minicpm-s25, hub-repeatability, minicpm-mobile-parity.
Next: Supply a qualified complete mobile checkpoint/export and validate the full multimodal/stateful pipeline.
Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy × InternVLA-A1
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Mobile Galaxy S24 / Snapdragon 8 Gen 3 × Spark-X2.5
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Mobile Galaxy S24 / Snapdragon 8 Gen 3 × Qwen3-TTS 0.6B CustomVoice
COMPONENT — C-historical. Historical FP16 decoder parity has SNR 39.13 dB. Tested calibrated INT8 decoder parity failed with negative SNR. Neither result is a full local TTS stream.
Evidence: tts-mobile-parity.
Next: Validate the complete local pipeline and a quality-passing artifact set.
Mobile Galaxy S24 / Snapdragon 8 Gen 3 × Qwen3.8-27B
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Mobile Galaxy S24 / Snapdragon 8 Gen 3 × MiniCPM-o 4.5
COMPONENT — C-historical. Historical FP16 speech-head job re-fetched successfully; SNR 49.22 dB/top-1 match in the recorded parity test. W8A16 top-1 failed. Not complete MiniCPM-o.
Evidence: minicpm-s24, minicpm-mobile-parity.
Next: Qualify thinker/encoders/vocoder and stateful local integration.
Mobile Galaxy S24 / Snapdragon 8 Gen 3 × InternVLA-A1
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Embedded SA8775P ADP / AI Hub × Spark-X2.5
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Embedded SA8775P ADP / AI Hub × Qwen3-TTS 0.6B CustomVoice
PROFILE_ONLY — C-timing-only. Historical NPU talker and GPU vocoder profiles re-fetched successfully; no complete pipeline, parity or co-residency qualification.
Evidence: tts-sa8775p-npu, tts-sa8775p-gpu, tts-other-devices, tts-other-device-code.
Next: Validate target-specific numerical quality, complete local execution, memory and sustained behavior.
Embedded SA8775P ADP / AI Hub × Qwen3.8-27B
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Embedded SA8775P ADP / AI Hub × MiniCPM-o 4.5
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Embedded SA8775P ADP / AI Hub × InternVLA-A1
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Embedded RB3 Gen 2 / QCS6490 / AI Hub × Spark-X2.5
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Embedded RB3 Gen 2 / QCS6490 / AI Hub × Qwen3-TTS 0.6B CustomVoice
PROFILE_ONLY — C-timing-only. Historical GPU vocoder profile re-fetched successfully. The attempted NPU talker build was rejected for floating-point inputs. No full model pipeline.
Evidence: tts-rb3-gpu, tts-rb3-talker-rejected, tts-other-devices, tts-other-device-code.
Next: Validate target-specific numerical quality, complete local execution, memory and sustained behavior.
Embedded RB3 Gen 2 / QCS6490 / AI Hub × Qwen3.8-27B
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Embedded RB3 Gen 2 / QCS6490 / AI Hub × MiniCPM-o 4.5
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.
Embedded RB3 Gen 2 / QCS6490 / AI Hub × InternVLA-A1
BLOCKED — none. No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device.
Evidence: hub-inventory, host-wsl, backend-factory, graph-contract.
Next: Supply and validate the missing matching artifact/backend, then run the complete specified workload.