Updated

Chord

Decagon introduced Chord, its first speech model, trained and post-trained for customer conversations and used by its Voice 3 system. The company says Chord shapes speech phrase by phrase and is trained on licensed data and consented voice talent; Voice 3 supports more than 70 languages.

By No Code MBA · Facts checked

Provider
Decagon
Announcement date
Status at last check
Generally available
Model family
Chord
Capabilities
Audio

What changed

Chord is Decagon's first in-house speech model, post-trained on real customer service conversations and trained on licensed data and consented voice talent, not customer-owned data. It adjusts pacing phrase by phrase, for example slowing down for confirmation codes or phone numbers. It powers Voice 3, a duplex agent that pairs a low-latency conversational model with a separate reasoning and tool-calling model and supports more than 70 languages. Decagon reports about 90% of listeners in a blind test could not tell Chord from a human.

Access and availability

Chord is used inside Decagon's Voice 3 agent on the Decagon customer service platform. Access is through Decagon sales (book a demo); no standalone API was announced.

Official access documentation

What you can build

If evaluating a customer-service voice stack, test real turn-taking, interruptions, confirmation details, pronunciation, and handoff behavior across representative calls.

Limitations

Chord is not offered as a standalone model or API. Decagon published no latency figures, and the blind listening test size and method were not disclosed.

Sources and release history