Models
One name per model, however many providers serve it.
Moondream: Parakeet Ultra
Accepts- Audio
Creates- Text
Parakeet Ultra is Moondream's post-trained version of NVIDIA's Parakeet TDT 0.6B v3 speech-to-text model, with the same architecture, tokenizer, and 0.6B parameters.
moondream/parakeet-ultraNov 3, 2026With segment and word times361 Toman a minute of audioMoondream: Moondream 3 Preview
Accepts- Text
- Image
Creates- Text
ReasoningMoondream 3 Preview is a vision-language model from Moondream with 9B total parameters and 2B active (mixture-of-experts).
moondream/moondream3-previewOct 15, 202632K context109,312 Toman/M input tokens901,824 Toman/M output tokensMoondream: Moondream 3.1
Accepts- Text
- Image
Creates- Text
ReasoningMoondream 3.1 is a vision-language model from Moondream, the successor to Moondream 3 Preview, with 9B total parameters and 2B active (mixture-of-experts).
moondream/moondream3.1-9b-a2bOct 15, 202632K context109,312 Toman/M input tokens901,824 Toman/M output tokens
3 of 581 models · Catalog updated 3 hours ago. · Free plan prices, in Toman at today's rate. Plans and prices

