Release10d ago
Block unveils Dessa: a full-duplex speech model trained from scratch only on human conversation audio, running locally on a MacBook Pro M4 Max
On 2026-10-23 (16:53 UTC) the official @blocks account unveiled Dessa, and Jack Dorsey called it the most authentic full-duplex speech model ever built. The official page block.xyz/dessa says Dessa is a true speech-to-speech model trained only on recordings of human conversation, with no text and no pretrained models. It is full-duplex, so it listens and talks at the same time, cuts in and backchannels naturally, runs faster than real time, and runs locally on a MacBook Pro M4 Max or better. Block says it was trained on a single NVIDIA GB200 NVL72 and learned language from hundreds of times fewer words than LLMs. There is no public API or weights yet; a small group of AI safety researchers is getting early access.
Benchmark49d ago
Artificial Analysis: GPT-Live-1 (Astra) tops Speech to Speech Index at 81.5
Artificial Analysis (@ArtificialAnlys) published its Speech to Speech Index: OpenAI GPT-Live-1 with Astra (medium) as the delegated backend debuts at #1 with 81.5, while Sol (low) scores 80.1 at #3, with Grok Voice Think Fast 2.0 High between them at 81.3. GPT-Live-1 is a full-duplex speech model that can delegate reasoning and tool use to a backend text model mid-conversation. The same drop covers Speech Agent Arena, Tau Voice, Big Bench Audio, plus cost and time-to-first-audio.
Release18h ago
Hume Launches Audio Decision API: Ask Plain-Language Questions of Raw Audio in One Call, 70.7% Agreement With a Human Panel per Hume
On November 2 voice AI company Hume launched the Audio Decision API, which it calls the first audio-native model that answers questions about speech and sound. Developers upload raw audio, write each question as a sentence with candidate answers, and get back a probability for each answer, with no transcription step. It supports yes/no, multiple-choice, rating-scale and time-span questions, such as how many speakers there are, whether someone laughs, whether a caller sounds frustrated, or whether a siren is in the background. Per Hume, it was trained on more than 10 million human judgments and in an initial evaluation reached 70.7% agreement with a human panel's average judgment, versus 56.5% for an individual rater. Access is currently by request or waitlist.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.