Moshi
Moshi is a speech-text foundation model and full-duplex spoken dialogue framework. It uses Mimi, a state-of-the-art streaming neural audio codec.
Moshi scores 5.5/10 (Solid) in Voice & Realtime.
Language Python License Apache-2.0 Stars 10.5k★ Type model Self-host Yes Repo kyutai-labs/moshi Homepage https://kyutai.org
Quality
Solid 5.5
Adoption 51
Activity 33
Maturity 88
Community 51
Capability 50
Show the math
Overall 5.5/10
value = 5.5/10
score = value
Adoption 5.5/10
value = 50.5/100
score = 1 + 9 * value / 100
Activity 3.9/10
value = 32.5/100
score = 1 + 9 * value / 100
Maturity 8.9/10
value = 87.6/100
score = 1 + 9 * value / 100
Community 5.6/10
value = 50.8/100
score = 1 + 9 * value / 100
Capability 5.5/10
value = 50/100
score = 1 + 9 * value / 100
Key metrics
10.5k★ GitHub stars
3/90d Recent commits
2 devs/90d Recent contributors
1.9y Project age
2mo ago Last commit
Gotchas
No gotchas documented yet. Contribute one if you know a constraint we missed.
Labels
- Self-hosted
Self-hosted
You deploy and operate it yourself; there is no hosted option here.
speech-to-speech full-duplex realtime low-latency foundation-model