AI Briefing
KOSign in

Marimba: Fast, Lightweight Speech Model with Open Weights and Typed Decisions

·2026.11.16 01:30

Key point

Marimba is a fast, lightweight speech model that returns typed decisions using a frozen Qwen3-0.6B backbone.

Details

  • Marimba is a fast, lightweight model for audio designed to return typed decisions such as yes/no, a choice, or a score.
  • The system runs on a frozen Qwen3-ASR-0.6B backbone, utilizing zero-shot scores from the ASR decoder's hypotheses for word and language questions followed by calibration.
  • Small readouts on the frozen encoder handle voice activity, turn endings, speaker changes, and speaker count.
  • Open weights are available under the Apache-2.0 license, with installation via pip install marimba-audio and a paper on arXiv.
open-sourceqwen3-asrspeech-recognitionaudio-aimarimba
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

RedditNetEase Youdao releases Confucius4-R2T2, a stable-prefix streaming ASR modelRedditSalesforce releases SOV-Transcribe-1, an open-weight speech recognition modelPyTorchKRllama.cpp v1.0.0, First Major Release in Three and a Half YearsGeekNewsGitHub Next releases Ace, a local collaborative coding environment for humans and AI agentsCloudflare BlogCloudflare open-sources Kitesurf, its stateless browser for AI agentsHacker NewsGnomon: Git-like version control for Postgres data built for AI agents