Canary 1B Flash

Name: Canary 1B Flash
Author: NVIDIA

NVIDIA's fast multilingual speech model for transcription and translation in English, German, French and Spanish.

Official page

Run it locally

Memory: Runs on consumer NVIDIA GPUs
Formats: Safetensors (Hugging Face)

Overview

Canary-1B-Flash is a lighter, faster version of NVIDIA’s Canary model, built on the FastConformer encoder with a Transformer decoder.

Highlights

883M parameters
Transcription in four languages
Speech translation to and from English
CC BY 4.0 license

Typical uses

Fast multilingual transcription and subtitles.

Things to consider

Best run with NVIDIA NeMo.

Specifications are taken from the model card on Hugging Face and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.

#Speech Models #Open Weights #NVIDIA

Similar models

View all

CohereMar 26, 2026

Cohere Transcribe

Cohere's open 2B-parameter speech-recognition model for 14 languages, released under Apache 2.0 and also offered through Cohere's API.

Type: Speech
Context: —
Weights: Open

SesameMar 13, 2025

Sesame CSM-1B

Sesame's Conversational Speech Model that generates speech conditioned on conversation context, for more natural turn-taking voices.

Type: Speech
Context: —
Weights: Open

BlockOct 23, 2026

Dessa

Block's speech-to-speech model, described as trained from scratch on human conversation alone; early access is limited to a small group of AI safety researchers.

Type: Speech
Context: —
Weights: Closed