Introduction

Sesame, the voice-AI company behind the Maya and Miles conversational voices, shipped its October release on October 6, 2026. The Sesame voice assistants are now available to everyone in the Sesame app on iOS and Android, running on what the company calls a new voice model: an updated version of its Conversational Speech Model (CSM) tuned for better expressivity and pronunciation.

This is a closed, app-only release. There are no new weights, but you can download the app and talk to it today. Sesame's only open checkpoint is still csm-1b, the base CSM it released on Hugging Face in March 2025.

Pro Tip

Judge this one over a long call, not a single line. Sesame says the CSM update makes "a huge difference, especially over longer conversations," so give it ten minutes of back-and-forth before deciding how it compares to your current voice stack.

What Shipped

From Sesame's October launch post and its getting-started guide:

  • Updated Conversational Speech Model: better expressivity and pronunciation, aimed at long, natural conversations.
  • Four characters: Maya (witty collaborator), Miles (laid-back, tells it like it is), Simone (deep-dive thought partner), and Charlie (casual sounding board). Each one has its own voice and personality, and you can call or text any of them.
  • Less sycophancy: Sesame says it reduced sycophancy and cut down on repetitive phrases, so the agents push back when you're wrong.
  • Its own computer: each agent can now browse the web and go off on longer research tasks, then report back with links, files, or voice notes when it's done.
  • Connectors: Gmail, Google Calendar, and Google Drive to start, plus your own external MCP servers that use OAuth.
  • Wider access: the voice assistants are "available to everyone today," on both the App Store and Google Play.

Voice as a Remote for Coding Agents

The most unusual part of the release is how Sesame says its own team works: talking to Maya, Miles, Charlie, and Simone to orchestrate other agents like Devin, Claude Code, and Codex, by voice, on walks and commutes.

That's arriving for everyone as Sesame Link, an early preview at link.sesame.com that the site describes as "your coding sessions, from your browser or a voice call," connecting Sesame to the agents running on your own computers. Sesame also says developer tooling for building your own Sesame apps is coming later this year.

The Bigger Plan: Glasses in 2027

All of this is a preview of the assistant Sesame wants to put on your face. The company is designing its assistant as a voice-first operating system for its own glasses, which are due in 2027 and will be Made in Japan, as it announced on October 5. Sesame's roadmap has three parts: frames people want to wear, assistance they actually use, and characters they enjoy talking to, which it develops with writers, producers, and actors working alongside its research team.

A Prompt to Try

The best way to hear what changed is to push the model on tone, timing, and pronunciation in one go. Open a call with any character and try something like this (my own test, not from Sesame):

I'm practicing a toast for my sister's wedding and I'm nervous. Listen to my
first line, then tell me honestly whether it sounds warm or stiff. After that,
read it back to me twice: once like a calm radio host, once like you're
holding back happy tears. Then say "Saoirse," "Nguyen," and "Worcestershire"
slowly, and use them in one sentence at normal speed.

You'll hear whether it reads the room, shifts delivery on request, and handles tricky names: the three things Sesame says it improved.

Why It Matters

Most voice launches compete on raw text-to-speech quality or latency. Sesame is betting on presence and character: voices that hold up over a long conversation and agents that can actually do work while you talk. Opening the app to Android and to everyone gives creators and developers a free, hands-on benchmark for what a conversational voice can sound like in 2026, even if the model itself stays closed.

Conclusion

Sesame's October 6 release brings an updated Conversational Speech Model to Maya, Miles, Simone, and Charlie, with better expressivity and pronunciation, plus agent computers, Google and MCP connectors, and a Sesame Link preview for steering coding agents by voice. It's live now in the Sesame app on iOS and Android, while the glasses it's all building toward arrive in 2027.

Primary: Sesame, "A new form of eyewear" (Oct 6, 2026), Sesame getting-started guide,

, Sesame Link preview.

—Anabel ♡