Open dataset · v1.3

MyAI Bench Contributors

Every benchmark is a contribution.

Thanks to these people, projects, and runtimes for keeping the data honest. The dataset is CC-BY-4.0 and every submission credits its author.

565
Records in v1.3
75
Unique devices
4
Recognized contributors
325
Records attributed

The roll call

Editorial, upstream runtimes, and the first community contributors. New names land here when a submission closes with the benchmark-accepted label.

  • MyAIHardware Editorial

    @myaihardware · Global

    Editorial team. Aggregates llama.cpp + vLLM logs, MLPerf submissions, and vendor numbers into the public dataset while we onboard community contributors.

    325 recordssince 2026-04-01Profile

  • llama.cpp upstream

    @ggerganov · Sofia, BG

    Many of our LLM numbers reference llama.cpp PRs and discussions. Credit where it's due — they wrote the runtime.

    since 2026-04-01Profile

  • vLLM project

    @vllm-project · Berkeley, CA

    Reference vLLM throughput numbers ground our datacenter-GPU rankings.

    since 2026-04-01Profile

  • Hugging Face TEI

    @huggingface · Paris, FR

    Text-embeddings-inference is our reference runtime for BGE-Large numbers.

    since 2026-04-01Profile
How to be listed here

Submit a number, get credited.

Run any of the 12 MyAI Bench workloads on your hardware. Submit via the form (it forwards to a public GitHub issue). When the issue closes with the benchmark-accepted label, your handle lands on this page next cut.

  1. 1Run a workload using the methodology, same prompts, same seeds, median of 5 runs.
  2. 2Use myai-bench (recommended) or any runtime, log a public source URL.
  3. 3Submit via /benchmarks/submit or open a GitHub issue with the benchmark-submission label.
  4. 4Editorial reviews + merges into the next bench cut. You get credit, your handle gets a card.
Submit a benchmarkSubmission queueAccepted benchmarks
New here? Back to MyAI Bench · See the v-by-v changelog