Kokoro v1.0

Training compute

1.7×10²⁰ FLOP

Parameters

82M

Published

Jan 27, 2025

Kokoro v1.0 is an AI model developed by hexgrad, first published in January 2025. It works in the speech domain, on tasks such as text-to-speech (tts) and speech synthesis.

Training it took an estimated 1.7×10²⁰ FLOP of compute (estimation method: hardware). The model has 82,000,000 parameters. Training ran on NVIDIA A100 SXM4 80 GB.

Access: Open weights (unrestricted). Its weights are openly available. Epoch AI rates the confidence of this record as confident.

Full record

Organization: hexgrad
Domain: Speech
Task: Text-to-speech (TTS), Speech synthesis
Training compute: 1.7×10²⁰ FLOP
Compute estimation method: Hardware
Parameters: 82,000,000
Training hardware: NVIDIA A100 SXM4 80 GB
Chip-hours: 500
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Epoch confidence: Confident