Of 28 public speech-to-text benchmarks, 17 carry a date — and 11 active pages expose none

Asked:

"List every public STT benchmark or leaderboard, what it measures, which vendors are on it and when it was last updated."

A broad web scan up to 2 September 2026 found 28 named public speech-to-text / ASR benchmark or leaderboard series: 9 live leaderboards, 13 recurring benchmark or report series, 1 dated report and 5 one-off public studies. Dated items run from October 2022 to 27 August 2026; 11 active pages expose no reliable update date and are reported as verified accessible on 2026-09-02, not as updated then.

Every benchmark by last update date, coloured by format

Live/public leaderboard Benchmark or report series Dated public report One-off public study Hollow ring = undated, page verified 2026-09-02

What the timeline cannot say

The freshest dated items are Pipecat (2026-08-27), Open Universal Arabic ASR (2026-08-08), Coval/Openbenchmarks and the Structured STT benchmark (2026-08-07), FFASR (2026-08-03), GigaSpeechBench (2026-07-28) and the Hugging Face/SpeechColab Open ASR Leaderboard (repository update 2026-07-13) — sources linked in the table below.

Many comparisons are vendor-published — AssemblyAI, Soniox, Speechmatics, Gladia, Picovoice, Voicegain and Corti all rank their own systems — so the catalog does not imply independence; see for example soniox.com and assemblyai.com.

The broadest multi-vendor boards are the huggingface.co Open ASR Leaderboard (23 vendors/projects across 11 datasets) and artificialanalysis.ai (29 vendors/projects, accuracy plus speed, latency and price).

Scope varies widely: medical speech alone has four separate benchmarks (Corti, Speechmatics, AssemblyAI, Omi Health), and language-specific boards cover Arabic, Korean, Danish and far-field room acoustics — e.g. benchmarklist.com.

The full catalog, 28 rows

Benchmark	Format	Publisher	Measures / scope	Metrics	Vendors / projects represented	Last update	Source

Catalog of publicly discoverable named comparative STT/ASR benchmarks and leaderboards: 28 rows found in a broad web scan with research cutoff 2026-09-02. "Every" means every named public comparison found — live boards, reproducible academic boards, vendor-published recurring comparisons, dated reports and notable one-off studies; it excludes bare datasets without posted results, single-model evaluations, TTS-only or speech-to-speech-only rankings, diarization-only tests and private evaluations. Web-wide completeness cannot be guaranteed. Dates are stated benchmark/repository update dates where available, otherwise page/report publication dates (see each row's date basis in the tooltip and last-update column); undated pages are reported as verified accessible on 2026-09-02, never as updated then. Blank cells mean the public source exposed no value. The vendors column mixes vendors with open-source projects/model families because the public sources do. Long "named systems" lists were trimmed to the vendors column for space.

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.