GitHub Repository

Very low latency speech to text, intent recognition, and text to speech, for building voice agents and interfaces

8,437 starsC

Moonshine Open-Weights STT models – higher accuracy than WhisperLargev3

Name: Moonshine Open-Weights STT models – higher accuracy than WhisperLargev3
Availability: InStock
Author: petewarden

by petewarden·Feb 24, 2026·316 points·81 comments

Visit Project View on HN

AI Analysis

●●●BangerShip ItSlick

Beats Whisper v3 accuracy on $100K budget; shipping on six platforms now.

Strengths

•Verified WER improvement over Whisper Large v3 (state-of-the-art competitor) on HuggingFace leaderboard
•Genuine multi-platform shipping: Python, iOS, Android, macOS, Linux, Raspberry Pi, wearables
•Streaming-first architecture with low-latency inference while user still speaking

Weaknesses

•Whisper v3 is 2+ years old; newer proprietary models (Deepgram, Llama Speech) likely exceed it
•No breakdown of WER by language, accent, or noise type; leaderboard ranking is aggregate
•Small team budget claim (sub-$100k/month) is context, not differentiation

Post Description

I wanted to share our new speech to text model, and the library to use them effectively. We're a small startup (six people, sub-$100k monthly GPU budget) so I'm proud of the work the team has done to create streaming STT models with lower word-error rates than OpenAI's largest Whisper model. Admittedly Large v3 is a couple of years old, but we're near the top the HF OpenASR leaderboard, even up against Nvidia's Parakeet family. Anyway, I'd love to get feedback on the models and software, and hear about what people might build with it.