Item detail
github.com

ONNX Runtime: MIT Microsoft Cross-Platform ML Inference + Training Accelerator (21,042*, ONNX Standard, CPU/GPU/NPU Hardware Acceleration, 100+ Models Supported)

RepoRadar surfaced ONNX Runtime: MIT Microsoft Cross-Platform ML Inference + Training Accelerator (21,042*, ONNX Standard, CPU/GPU/NPU Hardware Acceleration, 100+ Models Supported) — an AI project — into the Radar section, where it sits at Gold tier with a 'try now' verdict. Its strongest signal is workflow potential, scored 9.4 out of 10.

Score8.7
Popularity0.0
Risklow
TierGold
Score breakdown
Usefulness9.0
Novelty7.0
Momentum9.0
Maturity6.8
Open-source/build8.4
Evidence7.2
Workflow potential9.4
Setup ease6.4

Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.

Why it matters

Most AI / ML engineers + MLOps engineers + production ML platform teams deploying ML models today have been either (a) using a single-vendor inference stack (TensorFlow Serving / TorchServe / Triton / BentoML / Ray Serve) that locks-in the deployment model and lacks the canonical ONNX interchange format + cross-platform hardware accelerators, (b) hand-rolling a custom inference pipeline per

Who should use it

Builders Power users

Who should skip it

Pass on ONNX Runtime: MIT Microsoft Cross-Platform ML Inference + Training Accelerator (21,042*, ONNX Standard, CPU/GPU/NPU Hardware Acceleration, 100+ Models Supported) if its scope or audience does not match what your team is building right now.

About this signal

ONNX Runtime: MIT Microsoft Cross-Platform ML Inference + Training Accelerator (21,042*, ONNX Standard, CPU/GPU/NPU Hardware Acceleration, 100+ Models Supported) is tracked by RepoRadar as an AI project in the Radar section. First seen 2026-07-09; the source record was last checked on 2026-07-09. The current verdict is 'try now' with a Gold tier and moderate setup difficulty. The standout signals for ONNX Runtime: MIT Microsoft Cross-Platform ML Inference + Training Accelerator (21,042*, ONNX Standard, CPU/GPU/NPU Hardware Acceleration, 100+ Models Supported) are workflow potential (9.4) and practical usefulness (9.0), while setup ease (6.4) trails — that balance shapes where it fits best. This page summarizes the public evidence on the linked source page and states where additional review is still needed.

How this item is evaluated

The ONNX Runtime: MIT Microsoft Cross-Platform ML Inference + Training Accelerator (21,042*, ONNX Standard, CPU/GPU/NPU Hardware Acceleration, 100+ Models Supported) record combines a 8.7/10 composite score with separate popularity (0.0), risk (low), and setup (moderate) signals. See the scoring methodology for the current weights and evidence definitions.

Putting this into practice? Read How to evaluate an AI tool before you adopt it for the checklist behind this score.

Risk explanation

The 21; 042* repo is at production-grade maturity (MIT license; Microsoft official maintenance) but the consumer SHOULD note the execution provider selection (CPU / CUDA / TensorRT / DirectML / ROCm / OpenVINO / QNN / CoreML / WebGPU / NNAPI / XNNPACK) determines hardware support -- pick the right provider for your deployment; the consumer SHOULD review the quantization strategy before production deployment.

Evidence links
Closest alternatives / related signals
open-source mit onnxruntime microsoft onnx onnx-standard inference ml-inference
Verification record

What RepoRadar actually verified

Tested in a bounded workflow

Bounded representative workflow retained by RepoRadar verification harness. Last checked 2026-07-13T10:36:04.092395Z.

partial · cohort-20260712-onnxruntime-import

Tester
RepoRadar automated local verification harness
Started
2026-07-13T03:22:44.802107Z
Completed
2026-07-13T03:22:45.727152Z
Environment
Windows 10 AMD64; Python 3.11.9; credential-stripped child environment; disposable home/cache
Install/setup time
0 minute(s)
Evidence scope
Bounded setup or capability check
Cleanup
Per-check temporary home and work directory removed. Shared cohort package cache removed.
Actions exercised
  • Created a disposable home, work directory, and isolated package cache with credential-like environment variables excluded.
  • Executed bounded check: Install and import ONNX Runtime in an isolated Python environment.
  • Captured the complete sanitized stdout, stderr, exit status, and 0.92-second wall time.
Observed results
  • Command exited 0 after 0.92 seconds.
  • Expected marker 'CHECK_OK' was observed in retained output.
Observed strengths
  • onnxruntime==1.27.0 installed and imported successfully in the isolated Python environment, emitting the expected version marker.
Friction
  • The command emitted stderr; warnings or errors are preserved in the retained log for review.
Limitations
  • No ONNX model was supplied, so inference providers, performance, and model compatibility were not tested.
  • This bounded cohort check is not a production benchmark or a claim of real user-workflow adoption.

Pricing assessment: No paid plan or metered provider usage was exercised; package or licensing, hosting, and provider costs remain workflow-dependent.

Privacy assessment: No repository content, user data, or provider prompt was transmitted; broader product data handling was not assessed by this bounded run.

Open retained test log →

passed · cohort-20260712-onnxruntime-cpu-inference-workflow

Tester
RepoRadar automated local verification harness
Started
2026-07-13T10:35:57.334597Z
Completed
2026-07-13T10:36:04.092395Z
Environment
Windows 10 AMD64; Python 3.11.9; credential-stripped child environment; disposable home/cache
Install/setup time
1 minute(s)
Evidence scope
Bounded representative workflow
Cleanup
Per-check temporary home and work directory removed. Shared cohort package cache removed.
Actions exercised
  • Created a disposable home, work directory, and isolated package cache with credential-like environment variables excluded.
  • Created 1 synthetic fixture file(s) inside the disposable work directory; retained hashes prove the exact inputs.
  • Generated and serialized a two-input-shape ONNX Add graph with a fixed bias tensor and opset 13.
  • Loaded the model through ONNX Runtime's CPUExecutionProvider, ran a numeric input, and asserted the exact [3, 4] output.
  • Executed bounded check: Build a tiny ONNX add graph and execute it with ONNX Runtime's CPU provider.
  • Captured the complete sanitized stdout, stderr, exit status, artifact checks, and 6.76-second wall time.
Observed results
  • Command exited 0 after 6.76 seconds.
  • ONNX Runtime loaded the generated model and executed its Add node on CPU with the exact expected numeric output.
  • Expected marker 'CHECK_OK provider=CPUExecutionProvider output=3,4' was observed in retained output.
  • Validated model.onnx: 0 required marker(s) present and 0 excluded marker(s) absent; size and SHA-256 are retained.
  • Validated result.json: 2 required marker(s) present and 0 excluded marker(s) absent; size and SHA-256 are retained.
Observed strengths
  • The runtime accepted a standards-based serialized model and provided deterministic CPU inference through its normal InferenceSession API.
Friction
  • Creating a self-contained inference fixture required the separate ONNX graph-construction package and an explicit compatible IR/opset version.
  • Setup or runtime emitted 3 stderr line(s); the complete warnings/errors are preserved in the retained log.
Limitations
  • The generated two-value graph validates model loading and CPU inference only; it does not benchmark real model operators, accelerators, quantization, memory use, or serving concurrency.
  • This credential-free disposable workflow does not establish production scale, model quality, reliability under sustained use, or team adoption.

Pricing assessment: The graph build and CPU inference used local open-source packages with no model download, accelerator service, or hosted endpoint.

Privacy assessment: The generated model and numeric tensor remained inside the disposable work directory and local CPU process.

Open retained test log →

Verification sources

Longitudinal intelligence

How this decision record is moving

Raw history JSON →

28 dated snapshots retained from 2026-07-09 through 2026-08-13; see the snapshot index for explicit coverage gaps. Stars, version, release, pricing, integration, risk, maintenance, verdict, score, and momentum fields remain explicit even when a source has not reported them. Repository momentum is a normalized 0–10 RepoRadar signal; GitHub stars appear only where the popularity monitor retained exact timestamped observations.

RepoRadar score8.7 current · +0.0 net
Repository momentum9.3 current · +0.3 net
GitHub stars (observed)21,369 current · +284 net
GitHub stars21,369 exact observation
Versionv1.29.0
Last release2026-08-12T06:15:37Z
Maintenanceactive
Current risklow
Current verdicttry now
Pricing baselineNo structured commercial pricing baseline
Pricing checkedNot applicable or not recorded
Pricing freshnessNo dated commercial pricing review
Integrations baselineNo structured integrations recorded

Recent dated points

DateScoreMomentumStarsRiskVerdictMaintenance
2026-08-138.79.321,369lowtry nowactive
2026-08-128.79.621,360lowtry nowactive
2026-08-118.79.621,342lowtry nowactive
2026-08-108.79.621,329lowtry nowactive
2026-08-098.79.621,319lowtry nowactive
2026-08-088.79.321,300lowtry nowactive
2026-08-078.79.321,256lowtry nowactive
2026-08-068.79.0Not recordedlowtry nownot recorded
2026-08-058.79.0Not recordedlowtry nownot recorded
2026-08-048.79.321,256lowtry nowactive
2026-08-038.79.321,256lowtry nowactive
2026-08-028.79.321,253lowtry nowactive

Why the record changed

stars changed

Stars changed: 21360 → 21369.

stars changed

Stars changed: 21342 → 21360.

version changed

Version changed: plugin-ep-webgpu/v0.2.1 → v1.29.0.

stars changed

Stars changed: 21329 → 21342.

stars changed

Stars changed: 21319 → 21329.

stars changed

Stars changed: 21300 → 21319.

stars changed

Stars changed: 21256 → 21300.

stars changed

Stars changed: 21253 → 21256.

stars changed

Stars changed: 21246 → 21253.

stars changed

Source-observed stars changed: 21243 → 21246. This reports the retained observation delta and does not infer why the upstream change occurred.

stars changed

Source-observed stars changed: 21242 → 21243. This reports the retained observation delta and does not infer why the upstream change occurred.

stars changed

Stars changed: 21207 → 21219.