Downloads · 30 days
12
14% of all-time downloads
Vashisht7/noteechoes-english-voice-intent-action-qwen3-0.6b-mlx-8bit
noteechoes-english-voice-intent-action-qwen3-0.6b-mlx-8bit is a text generation model from Vashisht7. Use it when you need the model to write or continue text. It is set up for mlx. The card lists the license as apache-2.0.
This is the production MLX 8-bit release of the NoteEchoes English voice intent/action model. It converts a spoken transcript into either clean normalized text or a schema-v5 JSON action proposal for tasks, checklists…
Downloads · 30 days
12
14% of all-time downloads
All-time downloads
84
Public
Parameters
596M
645 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors633 MB · 98%
How the weights are stored.
U32596M · 100%
From the Hugging Face model README
This is the production MLX 8-bit release of the NoteEchoes English voice intent/action model. It converts a spoken transcript into either clean normalized text or a schema-v5 JSON action proposal for tasks, checklists, messages, reminders, notes, prompts, memory search, calendar items, email, cancel/no-op/clarification, and related capture intents.
The model proposes actions; it does not execute tools or claim that an action succeeded. The application must validate the JSON schema, enforce confirmation and safety policy, and execute any approved action outside the model.
Qwen/Qwen3-0.6B83a814988549ede2fb5bdbff28914b001fc58e449681960aeda9cc09351fd1e8behavior_report.json and
release-manifest.json in this repositoryStrict byte-for-byte output accuracy is intentionally reported separately from operational accuracy because safe, schema-valid wording variations can be semantically equivalent. Production promotion requires 100% of the operational JSON/schema/safety/routing gates.
Use the included chat template with thinking disabled. Provide the NoteEchoes
Core schema-v5 system instruction followed by the user's raw English transcript.
Generation is deterministic (temperature=0) with a release ceiling of 384 new
tokens.