Skip to content
Trending storyDeveloping

HojoAI releases open multilingual ASR model Hojo-ASR-Multi-V1 with 3.54 average WER

1 article1 sourcesince Oct 16Last article 2h ago ·

Overview

AISummary of 1 article

HojoAI releases Hojo-ASR-Multi-V1, an open multilingual speech recognition model that scored 3.54 average word error rate (WER) across five languages on a benchmark table, according to a post by Rohan Paul.

The model pairs an acoustic encoder and adapter with a Qwen3-4B LLM decoder.

The model is available on Hugging Face under the Apache 2.0 license and can be installed through the hojo-asr Python package. Rohan Paul says the package supports local inference and WAV, byte stream and SCP inputs. The reported WER figure comes from the benchmark table in that post; this overview does not verify it independently.

Written by AI from the articles below · updated Oct 16, 3:52 AM ET

Check the sources:

Latest developmentOct 16, 3:27 AM ET
Hojo-ASR-Multi-V1 posts 3.54 average WER across five languages

Article timeline

The articles in this story. Times are ET.

Oct 16
  1. Rohan PaulX
    Hojo-ASR-Multi-V1 posts 3.54 average WER across five languages

    AIRohan Paul says HojoAI's open multilingual speech recognition model, Hojo-ASR-Multi-V1, scored 3.54 average WER across five languages on a benchmark table. The model pairs an acoustic encoder and adapter with a Qwen3-4B LLM decoder, and is available on Hugging Face under Apache 2.0. The post also describes the package, which supports local inference and WAV, byte stream and SCP inputs.

    Image from @rohanpaul_ai's post

Heat trend

Current comparable heat
9
Peak comparable heat
10Oct 16, 4:00 AM ET
Change in past 24 hours
–

The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.