Sesame Review, Pricing & Alternatives (2026)

Sesame is a consumer voice-companion company founded by Brendan Iribe, former Oculus co-founder and CEO, and Ankit Kumar, formerly CTO of Ubiquity6. It emerged publicly in February 2025 with Maya and Miles, then announced a $250 million Series B in October 2025. Sequoia said the early demonstrations reached more than one million people and produced more than five million conversation minutes. Sesame’s current product is a free iOS and Android app offering Maya, Miles, Simone, and Charlie through voice or text. The company is not positioned as a conventional developer speech platform. Its app is the main consumer offering, while CSM-1B is a separate downloadable speech model for developers assembling their own systems. Sesame says glasses are coming in 2027, extending its voice-first product direction beyond phones.

The Sesame app centers on direct conversations with four named assistants: Maya, Miles, Simone, and Charlie. Each supports voice and text interaction on iOS and Android, and the beta announcement described abilities to search, text, and think. A typical workflow begins with choosing an assistant, then speaking or typing while the system manages conversational turns. Sesame’s research treats timing, pauses, interruptions, rhythm, emotion, and personality as core parts of interaction, rather than adding speech after a standard text response. The app is the complete consumer product and is free to download, although Sesame does not publish official prices or quotas for its Standard and Max Intelligence levels. CSM-1B is separate: developers must combine it with a language model, audio components, and supporting software to build an application.

CSM-1B is a one-billion-parameter conversational speech model released in March 2025 and integrated into Hugging Face Transformers in May 2025. It converts text and audio inputs into RVQ audio codes through a Llama-style backbone, an audio decoder, and the Mimi codec. The model can use conversational context and supports batched inference, making it suitable for developers assembling speech pipelines rather than complete assistant applications. The released checkpoint is a base model, not the fine-tuned Maya voice, and it cannot generate text. Sesame’s repository recommends a CUDA-compatible GPU and access to Llama-3.2-1B alongside CSM-1B. English is the practical target, since the project says non-English performance is likely weak. CSM-1B is open source under Apache License 2.0, while the consumer Sesame app is not an open-source application.

The Sesame app is free to download on iOS and Android. Its current release notes mention standard and Max Intelligence access, but Sesame does not publish official prices or quotas for either level. The company also does not list a standard paid developer plan, public API rate, or model-hosting price for the consumer service. CSM-1B is a free downloadable speech model under Apache License 2.0, though running it requires a CUDA-compatible GPU, Llama-3.2-1B, Mimi, and supporting software. Self-hosting therefore adds hardware, software, and engineering costs. CSM-1B generates speech rather than text or a complete assistant. A separate hosted CSM-1B service has independent pricing that does not represent Sesame’s consumer rates. Treat the app’s free listing as the entry point, then check the current listing for any paid cadence or feature pricing.

Sesame’s strongest case is expressive companion interaction, not documented enterprise deployment. Its iOS and Android app offers Maya, Miles, Simone, and Charlie through voice or text, while CSM-1B is a downloadable speech generator that requires a language model and supporting components. The repository prohibits impersonation, fraud, deceptive content, and harmful or illegal use. Sesame’s terms say logged-in user content may be used to provide, improve, develop, and train its services, with an opt-out applying to content collected afterward. Compared with Character.AI and Pi, Sesame combines a consumer companion with a public speech-model release. Compared with ChatGPT Voice, it provides a smaller character-led product with less published developer detail. Compared with ElevenLabs, it is less suitable for a conventional commercial speech API. Choose Sesame for consumer voice companions or speech research, not enterprise API procurement.