Sesame CSM
Sesame
Key facts · verified Oct 13, 2026
Sesame CSM is a Sesame audio generation model released on 13 Mar 2025. Pricing: Open weights — free to self-host. Source: huggingface.co.
Sesame's CSM (Conversational Speech Model), an open-weight speech generation model under Apache 2.0 that generates audio codes from text and audio context, using a Llama backbone and a smaller audio decoder. It is a base model that produces a variety of voices but is not tuned to any specific voice, and Sesame recommends supplying conversational audio as context.
Pricing & specs checkedSource: vendor documentation · checked Oct 13, 2026LMArena: 1180 Elo (last listed Aug 22, 2026) · source · retrieved Aug 22, 2026
Key Features
Audio Generation Text-to-Speech
Full Specifications
Open weights — free to self-host | |
|---|---|
| Identity | |
| Developer | Sesame |
| Released | Mar 2025 |
| Licence | Apache 2.0 |
| Self-hostable | Yes — weights at sesame/csm-1b |
| Capacity | |
| Sample rate | 24 kHz |
| Capability | |
| Languages | English; limited non-English capacity |
| Rights | |
| Commercial licence | Yes — Apache 2.0 |
| Measured quality | |
| LMArena Elo | 1180 (checked Aug 2026) |
| LMArena rank | Rank 11 |