Simba 3.2 is live on the TTS API as the recommended English model
Simba 3.2 is now available on POST /v1/audio/speech and /v1/audio/stream — a streaming-native Simba 3 model with lower TTFB, richer expressivity, and a curated voice allow-list, recommended for new English integrations.
simba-3.2 is now available on POST /v1/audio/speech and POST /v1/audio/stream through the model field. It is the go-forward Simba 3 model for English: streaming-native, with lower time-to-first-byte and richer expressivity than simba-3.0, and it is what we recommend for new English integrations.
Simba 3.2 serves from a curated voice allow-list, so you pass one of its registered voice IDs as voice_id. It is English-only at launch — a non-English voice returns 400 with the rejected locale called out — and multilingual support is planned under the same model name, so no migration will be needed when it lands.
The Build changelog entry lists the launch voice IDs and the exact request shape.
SSML on Simba 3
Shipped the same day: <break> and <prosody rate> SSML tags are now honored on the Simba 3 models (simba-3.0 and simba-3.2), so pauses and speaking-rate control behave the same as on the Simba 1.6 models. The SSML guide has the full tag reference.