moondream
Moondream: Parakeet Ultra
moondream/parakeet-ultra
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
Moondream Parakeet Ultra is an audio-to-transcription model listed in the OpenRouter catalog and documented as a post-trained variant of NVIDIA's parakeet-tdt-0.6b-v3, retaining the same architecture, tokenizer, parameter count, output conventions and CC-BY-4.0 license. The model documentation reports lower word-error rates than that base checkpoint across English, multilingual, noisy, business and long-form speech tests, plus higher throughput when served through Photon on an NVIDIA GPU. Those results are source-reported and should be validated on target audio and latency requirements.[1][2]
Strengths
- Documented to accept audio and return transcription, with optional segment-level and word-level timestamps in Photon.[1][2]
- Model-card measurements report lower word-error rates than the original checkpoint across English ASR, 25-language FLEURS, business speech, background-noise and long-form evaluations.[2]
- The card describes a built-in voice-activity head that lets Photon split long recordings at pauses without an external VAD model, and reports higher throughput under its stated NVIDIA GPU setup.[2]
Limitations
- The evidence pack contains the OpenRouter listing and Moondream-hosted documentation, but no independent evaluation; reported accuracy and speed need validation on the intended workload.[1][2]
- The OpenRouter entry leaves context length unclear and lists no supported request parameters or benchmark data, so these details remain unverified rather than proven absent.[1]
- Published results cover specific datasets, languages, noise conditions and a Photon/GPU serving setup; they do not establish performance for other accents, domains, audio quality or deployments.[2]
Where to use it
- Worth evaluating for transcription pipelines that need word or segment timestamps, such as captioning or searchable meeting records.[2]
- A plausible candidate for multilingual speech-to-text across the documented 25-language set, after testing the target languages and accents.[2]
- May suit long recordings, calls, podcasts or noisy speech given the reported long-form, business and noise evaluations; verify results on representative audio.[2]
Sources
- [1] OpenRouter catalog · Moondream: Parakeet UltraSources checked: Nov 13, 2026
- [2] Model documentation · Moondream: Parakeet UltraSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
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. It lowers word error rate on the Open...
Specifications
- Model ID
- moondream/parakeet-ultra
- Canonical ID
- moondream/parakeet-ultra-20261102
- Creator
- moondream
- Added to OpenRouter
- 03.11.2026
- Context length
- 0
- Modality
- audio->transcription
- Input modalities
- audio
- Output modalities
- transcription
- Token pricing
- No per-token price in the source snapshot — see OpenRouter for the billing unit
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | — |
| Coding | No measurement | — |
| Agents | No measurement | — |
| Writing | No measurement | — |
| Reasoning | No measurement | — |
| Document analysis | No measurement | — |
| Vision | No measurement | — |
| Image generation | No measurement | — |
| Audio | No measurement | input: audio |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
No evaluations are supplied for this model.
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| PrimeIntellectbf16 | 0 | — | — | — | — | — |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.