ONNX Runtime (+ Web)
What this tool produced in the local studio, where it ran and what it cost.
Connect to my local studio
Optional. This page never contacts a local service on its own: pressing the button sends one health request to your own machine, and nothing else.
Identity
- Version
- 1.30.0 (pipeline/uv.lock)
- Released
- 2026-09-10
- Licence
- MIT
- Licence class
- attribution
- Ring
- Adopt
- Environment
- pipeline, accel, web
- Capability probe
- pass 2026-10-11 · NVIDIA RTX 5000 Ada Generation Laptop GPU · driver 595.95
- gemm_fp16_4096_tflops: 104
- gemm_bf16_4096_tflops: 107.4
- ort_cuda_conv_x50_s: 0.0271
- ort_vs_torch_max_abs_err: 3.5762786865234375e-7
- torch_max_vram_mb: 136
Role in mining
Parity reference, GPU baseline and live in-browser models.
Artefacts it produced
Not yet run
Provenance
Each artefact card carries its run, recipe, commit, GPU, driver, timings and throttle shares. Licence and data sources:
No artefact yet, so no provenance to show.
Where it ran
Limits and failures
No limit or failed variant recorded.
Reproduce
No published run of this tool yet.
Alternatives rejected
None recorded.