new reportFrom Walkie-Talkie to Telephone — the state of full-duplex voice AI, H1 2026
Fullduplex · the speech-to-speech observatory

Voice AI, measured in the open.

The state of speech-to-speech, full-duplex, and audio foundation models — latency, releases, corpora, and the benchmarks that keep score. Curated by hand, every claim linked to a primary source.

Get the latest full-duplex AI report.

One email a week from the observatory — models, benchmarks, datasets, and what moved.

what brings you here?
36
models & platforms
19
native full-duplex
45
benchmarks
1.3M h
corpus hours catalogued
02 · cadence

Model releases per half-year

When the 36 tracked models and platforms first shipped.

4
1
7
7
8
9
23·H2
24·H1
24·H2
25·H1
25·H2
26·H1
04 · benchmarks

Recently updated benchmarks

The scoreboards that moved last. Full landscape on the benchmarks desk.

benchmarkupdated
SpeechJBB2026-06
LALM Jailbreak Taxonomy & Eval2026-05
AIA — Acoustic Interference Attack2026-05
Full-Duplex-Bench v32026-04
Artificial Analysis Speech Arena2026-04
Big Bench Audio2026-03
03 · corpora

Largest catalogued corpora, in hours

Log scale — the pretraining giants and the interaction sets live orders of magnitude apart.

the four desks

Latest dispatches

all posts →
the reply channel

Live on Discord, weekly in your inbox.

Fullduplex is, as the name implies, a two-way channel. Lab news and show & tell in real time on Discord; the weekly report by email; new entries and corrections through the community board.

Get the latest full-duplex AI report.

One email a week from the observatory — models, benchmarks, datasets, and what moved.

what brings you here?