Eleven v3 leaves alpha with improved number pronunciation

ElevenLabs made Eleven v3 generally available on February 2, 2026. The text-to-speech model left alpha with changes to stability and the interpretation of numbers, symbols, and specialized notation.

Reading a number in context

A phone number, sports score, currency amount, and chemical formula all require different spoken forms. ElevenLabs highlighted improvements in deciding how those inputs should sound rather than simply reading each symbol in isolation.

The announcement gave examples of digit-by-digit phone numbers, scores read as a comparison, and currency amounts with the correct magnitude. It reported its own test results across 27 categories and eight languages.

Availability beyond alpha

Eleven v3 became generally available across ElevenLabs platforms. The release brought the updated model into ordinary speech-generation workflows, with pronunciation of specialized input still a practical part of evaluating the output.