Skip to page content
SAVRN
Search Plan a project

Open-weight model · Text generation

OzyAssist-Voice

by Angel lopez Xangel0s/OzyAssist-Voice

OzyAssist-Voice is an open-weight model for text generation from Angel lopez, released under Apache License 2.0. Its published files total 986.1 MB.

OzyAssist Voice es un modelo de inteligencia artificial de ultra-baja latencia (30+ tokens/segundo, tiempo de respuesta <330 ms), destilado y afinado específicamente para asistentes de voz interactivos, cowork en vivo, livestreaming y control de Windows.

Parameters—
Context—
Weights986.0 MB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads—

Model Card

By Angel lopez, published under apache-2.0, revision 5e980608dba3.

OzyAssist Voice es un modelo de inteligencia artificial de ultra-baja latencia (30+ tokens/segundo, tiempo de respuesta <330 ms), destilado y afinado específicamente para asistentes de voz interactivos, cowork en vivo, livestreaming y control de Windows. Diseñado bajo la arquitectura Zero-Docker de OzyAssist, este modelo corre localmente consumiendo apenas ~1.37 GB de VRAM, lo que permite operarlo en cualquier laptop o GPU mientras juegas o transmites en directo. Pruebas empíricas ejecutadas en el mismo entorno de hardware (NVIDIA Quadro RTX 4000 Max-Q): Descarga el modelo GGUF y ejecútalo con aceleración GPU: Crea un archivo Modelfile: Luego construye y corre el modelo: En el proyecto…

Read Angel lopez's full model card

OzyAssist Voice (1.5B Distilled GGUF)

OzyAssist Voice es un modelo de inteligencia artificial de ultra-baja latencia (30+ tokens/segundo, tiempo de respuesta <330 ms), destilado y afinado específicamente para asistentes de voz interactivos, cowork en vivo, livestreaming y control de Windows.

Diseñado bajo la arquitectura Zero-Docker de OzyAssist, este modelo corre localmente consumiendo apenas ~1.37 GB de VRAM, lo que permite operarlo en cualquier laptop o GPU mientras juegas o transmites en directo.


Especificaciones del Modelo

Característica Detalle
Arquitectura Base Qwen2.5-Coder-1.5B-Instruct
Técnica Destilación de Conocimiento + Fine-tuning LoRA 4-bit (Unsloth)
Formato GGUF Cuantizado (Q4_K_M)
Tamaño de Archivo 986 MB
VRAM Requerida ~1.37 GB (Contexto 16k con FlashAttention y KV Q8_0)
Velocidad de Inferencia ~30.0 - 32.5 tokens/segundo (NVIDIA RTX)
Latencia Primer Token / Frase ~330 ms (Ideal para síntesis por voz fluida Piper / SAPI)
Plantilla de Chat ChatML (<|im_start|>, <|im_end|>)
Idioma Principal Español (con soporte nativo para comandos en inglés y código)

Benchmark: Comparativa de Latencia y Velocidad

Pruebas empíricas ejecutadas en el mismo entorno de hardware (NVIDIA Quadro RTX 4000 Max-Q):

Tarea / Comando OzyAssist-Voice (1.5B) Modelo Estándar (7B) Mejora
Confirmación de Voz ("¿Estás listo?") 0.33s (333 ms) 1.75s 5.3x más rápido
Reacción Livestream (Saludo en vivo) 0.72s 2.77s 3.8x más rápido
Control de Apps ("Abre calculadora") 0.43s 1.74s 4.0x más rápido
Cerrar Ventana ("Cierra notepad") 0.41s 1.75s 4.3x más rápido
Ajuste de Audio ("Sube volumen 80%") 0.54s 1.08s 2.0x más rápido
Consumo de VRAM 1.37 GB 4.90 GB 3.6x menos VRAM

Cómo Usarlo

1. Con Llama.cpp / Llama-Server

Descarga el modelo GGUF y ejecútalo con aceleración GPU:

llama-server -m OzyAssist-Voice-1.5B-v3-q4_k_m.gguf -ngl 99 -c 16384 -fa on --port 8080

2. Con Ollama

Crea un archivo Modelfile:

FROM ./OzyAssist-Voice-1.5B-v3-q4_k_m.gguf
PARAMETER temperature 0.2
PARAMETER stop "<|im_end|>"
PARAMETER stop "<|im_start|>"
SYSTEM "Eres OzyAssist Voice, un asistente de voz y escritorio para Windows de ultra-baja latencia."

Luego construye y corre el modelo:

ollama create ozy-voice -f Modelfile
ollama run ozy-voice

3. En OzyAssist (Modo Turbo / Voz)

En el proyecto OzyAssist, puedes iniciar directamente este modelo con el script integrado:

powershell -ExecutionPolicy Bypass -File .\scripts
un_advanced_brain.ps1 -Turbo

O iniciar la consola interactiva de voz con soporte de audio en tiempo real:

cd backend
go run cmd/ozy/main.go voice

Capacidades y Casos de Uso

  • Asistentes de Voz en Tiempo Real: Ideal para integración con motores STT (Whisper/Vosk) y TTS (Piper TTS/Windows SAPI) con Barge-in instantáneo.
  • Livestreaming y Bots de Chat: Respuestas inmediatas al chat de Twitch/YouTube sin pausas que rompan el ritmo de la transmisión.
  • Automatización de Escritorio: Comandos rápidos para abrir/cerrar ventanas, mutear o ajustar volumen, telemetría de GPU/CPU y atajos en Windows.
  • Decodificación Especulativa (Speculative Decoding): Puede utilizarse como modelo borrador (draft model) para acelerar al modelo maestro de 7B (Xangel0s/OzyAssist-7B) al doble de velocidad.

Licencia

Este modelo se distribuye bajo la licencia Apache 2.0.

Identity and Version

Repository
Xangel0s/OzyAssist-Voice
Publisher
Angel lopez
Task
Text generation
Modality
Text
Library
Not stated by the source
Parameters
Not stated by the source
Languages
es, en
Revision
5e980608dba3a0b0253a62367e64fb13a9edf080
First published
2026-09-20
Last updated
2026-09-20

Files and Weights

4 files, 986.1 MB in total. The weights are 1 file totalling 986.0 MB in gguf.

Weights1 file · 986.0 MB
Documentation1 file · 4.3 KB
Other1 file · 568 B
Repository1 file · 1.6 KB
Every file
FileTypeSizeSHA-256
OzyAssist-Voice-1.5B-v3-q4_k_m.ggufWeights986.0 MB af704c8e8393
README.mdDocumentation4.3 KB —
ModelfileOther568 B —
.gitattributesRepository1.6 KB —

License and Download

License
apache-2.0
Access
Open weights, no gate
Download size
986.0 MB
Download from Angel lopez

Released by Angel lopez through its official repository on Hugging Face. Read the license.

Built From

  • Derived from Qwen/Qwen2.5-Coder-1.5B-Instruct
  • Quantized from Qwen/Qwen2.5-Coder-1.5B-Instruct

Memory Requirements

PrecisionWeights in memory
As published986.0 MB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About OzyAssist-Voice

Can I use OzyAssist-Voice commercially?

Yes. OzyAssist-Voice is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

Similar Models

Model · Text generation

opt-125m

AI at Meta

OPT was first introduced in Open Pre-trained Transformer Language Models and first released in metaseq's repository on May 3rd 2022 by Meta AI. Disclaimer: The team releasing OPT wrote an official model card, which is available in Appendix D of the paper. Content from this model card has been written by the Hugging Face team. To quote the first two paragraphs of the official paper OPT was predominantly pretrained with English text, but a small amount of non-English data is still present within the training corpus via CommonCrawl. The model was pretrained using a causal language modeling (CLM) objective. OPT belongs to the same family of decoder-only models like GPT-3. As such, it was…

Open weights other 2,048 tokens transformers

Model · Text generation

Ornith-1.5-9B-GGUF

Ornith

Chirp Chirp! We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement. Ornith-1.5 extends Ornith-1.0 (which was developed on top of Qwen3.5 and Gemma4 with additional continued pretraining, mid-training, and post-training) by expanding the self-improvement loop from scaffold and rollout optimization to jointly optimizing task generation, scaffold construction, and solution rollouts. Rather than relying on a fixed set of human-curated tasks and manually designed harnesses, Ornith-1.5 continuously generates new training tasks, discovers effective strategies for solving them, and improves the policy through reinforcement learning. For…

Open weights mit transformers

Model · Text generation

Ornith-1.5-35B-A3B-GGUF

Ornith

Chirp Chirp! We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement. Ornith-1.5 extends Ornith-1.0 (which was developed on top of Qwen3.5 and Gemma4 with additional continued pretraining, mid-training, and post-training) by expanding the self-improvement loop from scaffold and rollout optimization to jointly optimizing task generation, scaffold construction, and solution rollouts. Rather than relying on a fixed set of human-curated tasks and manually designed harnesses, Ornith-1.5 continuously generates new training tasks, discovers effective strategies for solving them, and improves the policy through reinforcement learning. For…

Open weights mit transformers

Model · Text generation

Ternary-Bonsai-2-27B-gguf

Prism ML

Full 27B-class reasoning in ternary transformer weights, for llama.cpp (CUDA, Metal, CPU) - \~5.9 GB language model (down from \~54 GB FP16) — full 27B-class reasoning on a standard laptop or a single GPU - 98.2% of FP16 intelligence retained: 84.78 average across 14 thinking-mode benchmarks — far above the conventional IQ2XXS build (72.59) at about 82% of its footprint, and within 0.4 points of UD-Q4KXL at three times the footprint - Retains thinking, reasoning, and agentic behavior deep in the sub-4-bit regime, where conventional low-bit representations collapse: math within half a point of full precision (96.57), coding level with the baseline (89.42), agentic tool calling at 74.92…

Open weights apache-2.0 llama.cpp

Uncensored Qwen3.8-27B, published as GGUF quantizations with the multi token prediction (MTP) head retained and verified. Refusal behaviour has been substantially reduced, not eliminated. See Measured behaviour for the numbers. Capabilities, training data, and architecture are otherwise unchanged. - Refusal directions removed with Heretic, which co minimizes refusal count against KL divergence from the base model. No handwritten refusal removal code, no finetuning, no additional training data. - Abliteration runs at bf16 (no 4 bit quantization). the resulting LoRA is merged into the bf16 base, so the published weights are not a quantized round trip. - mtp. tensors are copied verbatim from…

Open weights apache-2.0 llama.cpp

Model · Text generation

Ornith-1.0-9B-GGUF

Ornith

Aloha! Today, we are releasing Ornith-1.0, a self-improving family of open-source models for agentic coding. This model card documents Ornith-1.0-9B, the most lightweight member of the Ornith family, designed for efficient single-GPU deployment. Ornith-1.0-9B is a dense ~9B model (≈19 GB in bf16), so it serves comfortably on a single 80GB GPU. The recipes below stand up an OpenAI-compatible server; add --tensor-parallel-size / --tp if you want to shard across more GPUs. For a quick local test (or to script offline generation), load the model directly with Transformers. Make sure you have a recent release installed — see the Transformers installation guide; Ornith-1.0-9B requires…

Open weights mit transformers