AI Judge changed title after evaluation, original title: "Cartesia founder Karan Goel launches Sonic-3.5 and Ink-2 streaming models, claiming top benchmarks in TTS and STT"
Cartesia's latest releases target the demanding requirements of live voice agents, where Sonic-3.5 handles text-to-speech and Ink-2 manages speech-to-text. Both models post leading scores on Artificial Analysis streaming leaderboards from May 2026 data, with gains focused on endpoint detection and background noise resilience rather than raw quality alone.
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
Semantic endpointing and native turn detection reduce the need for external voice activity detection, while noise robustness helps maintain clarity during real conversations. These changes align directly with voice-agent workflows that demand sub-second responses.
The models are generally available now through Cartesia's WebSocket and REST endpoints plus the public playground, with usage-based pricing that includes monthly minutes or hours depending on the tier. Exact per-model rates for Ink-2 remain bundled rather than itemized.
Many users expressed excitement over Cartesia's Sonic-3.5 and Ink-2 releases because the models deliver highly natural low-latency streaming TTS and STT that unlock better voice agents.
If you switch to Cartesia from your existing provider, three months are on us.
Hear the difference today. 👇🏻
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
Within the span of a week, we launched streaming TTS (text-to-speech) and STT (speech-to-text) models that topped the leaderboards. I'm incredibly proud of the research team for their relentless pursuit of improvement, which have unlocked new state-of-the-art audio models on the Pareto frontier of speed and quality.
As a research problem, speech requires fusing both text and audio and is the gateway to general multimodal models. We built Sonic-3.5 and Ink-2 from the ground up, developing multiple innovations along the way in a direction that will scale to general real-time intelligence. I've personally been deeply involved in building these models and more; it's been a blast working with the incredibly talented research team here @cartesia, and I can't wait to show the world what's coming next :)
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
Two new models just dropped 👀
Sonic-3.5 and Ink-2 are the #1 streaming models for text to speech and speech to text
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
We’ve seen incredible traction this year, with enterprises transforming every part of their operations with Voice AI. We’re sharing what we’ve learned and are giving away a guide to the top 10 enterprise use cases for Voice AI.
Retweet and comment "CARTESIA" on the main video and we'll send you the guide and $100 in credits so you can get started today.
If you switch to Cartesia from your existing provider, three months are on us.
Hear the difference today. 👇🏻
There's no way call centers stay in business after this.
Listen to this conversation. You cannot tell I'm speaking to an AI voice agent here.
This is Cartesia's model. Right now, #1 in the world.
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
congrats on the release!
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
Insane to watch my friends at Cartesia move from launch to technology leader in the competitive speech synthesis/recognition market.
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
@krandiash Making the word more magical, one model at a time.
Go Cartesia!
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
@chongz
@cartesia mfw when i use the best model for STT and TTS and also touch a van de graaff generator
@krandiash mark the date. this is elevenlabs' openai-getting-mogged-by-claude moment.
@krandiash Say goodbye to 24/7 support
@krandiash Lol 11labs is so cooked, their speed is nothing compared to this
@_albertgu i'd like everyone to know i was also involved
played around with building a voice agent on @cartesia this morning.
a few things stood out for anyone running support or booking lines.
it waits until you've actually finished talking before it responds.
turn detection is built into the model so it stops cutting people off mid sentence the way most agents do.
it reads phone numbers and confirmation codes back cleanly which is usually where support bots embarrass themselves.
> 100ms to first audio means no awkward gap before it answers. very hard to tell apart from human speed.
the whole stack is two models, sonic 3.5 for speech and ink 2 for transcription, both ranked #1 in their category on artificial analysis right now.
if you've got customers stuck on hold or stuck repeating their account number, worth checking them out 👇
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
@krandiash I usually skip demos, but this one kept my attention.
@krandiash The future isn't chat. It's conversation.
@krandiash Is he whispering because he can't afford business class?
@krandiash Use this comment along with the GIF
@krandiash can't wait to fly on Air Cartesia 😍
