Crowdsourced data

Submit your benchmark

Got numbers from your rig? Submit them here. Every submission is reviewed, credited, and merged into the public dataset.

Submissions are forwarded to our public GitHub issue tracker, so you can see the queue and the review thread in the open. Fields marked with an asterisk are required.

Who are you?

So we can credit your run.

Name / handleEmail

What did you run?

Be specific, exact card name, exact quant, exact context.

Device (e.g. RTX 4090, Apple M4 Max, MI300X)WorkloadQuantizationContext length (tokens)Batch sizeVRAM (GB, or 0 if shared)TDP (W, measured at wall preferred)

What did you measure?

Fill at least one, the one your workload reports.

Tokens / sec (LLM)Images / min (SDXL)Real-time × (Whisper)

Where can we verify?

A link to your log / tweet / Reddit thread / Discord post. Required.

Source URL (gist, Reddit, X, GitHub issue)Runtime command (optional but encouraged)Notes (driver version, OS, ambient temp, anything weird)

Read the methodology first →