# Fast inference

Qwen3-0.6B-8bit is an 8-bit quantized version converted from Qwen/Qwen3-0.6B, a text generation model suitable for the MLX framework.

A lightweight and fast modified version of the FRIDA model for computing embedding vectors of Russian and English sentences

Lite-Whisper is a lightweight version of OpenAI Whisper compressed using LiteASR technology, significantly reducing model size while maintaining high recognition accuracy.

Kokoro is an open-source text-to-speech model with 82 million parameters, achieving sound quality comparable to large models with a lightweight architecture while improving generation speed and reducing computational costs.

This is an auto-trained image classification model capable of recognizing multiple common object categories

FP8 quantized version of Stable Diffusion 3.5 Large for text-to-image generation tasks.

Sana is an efficient text-to-image framework capable of generating images with resolutions up to 4096×4096, supporting multilingual input.

DepthPro is a foundational model for zero-shot metric monocular depth estimation, capable of generating high-resolution, high-precision depth maps.

This is an OpenVINO-optimized midjourney-mini model for text-to-image generation tasks.

Sana is an efficient text-to-image framework capable of generating images up to 4096×4096 resolution, deployable on laptop GPUs.

An efficient text-to-image model series distilled from mainstream diffusion models on AMD Instinct™ GPUs

A dual-framework compatible vision model trained on MetaCLIP-2.5B dataset, supporting both OpenCLIP and timm frameworks

A dual-purpose vision-language model trained on the MetaCLIP-2.5B dataset, supporting zero-shot image classification tasks

A dual-framework compatible vision model trained on MetaCLIP-2.5B dataset, supporting zero-shot image classification tasks

Vision Transformer model trained on MetaCLIP-400M dataset, supporting zero-shot image classification tasks

A dual-framework compatible vision model trained on the MetaCLIP-2.5B dataset, supporting both OpenCLIP and timm frameworks

A vision Transformer model trained on the MetaCLIP-2.5B dataset, compatible with both open_clip and timm frameworks

The 4-bit quantized version of Molmo-7B-O, significantly reducing the memory requirement and suitable for environments with limited resources.

A versatile vision+text generation model, particularly suitable for generating realistic images that simulate color film photography, presenting a wide range of visual paradigms from Autochrome to Kodachrome to Fujifilm and other iconic photographic technologies.

A 4-bit quantized text-to-image generation model optimized for the MLX framework, supporting efficient image generation

A zero-shot classification model based on the Transformers library, capable of performing classification tasks without task-specific training data.

A distilled version of ProtGPT2, compressed into a more efficient small model through knowledge distillation, maintaining performance while improving inference speed

The 4-bit quantized version of the Llama 3 model, which optimizes memory usage and speeds up inference, suitable for environments with limited computing resources.

An image classification model fine-tuned on the cats_vs_dogs dataset using Google's ViT model, designed to distinguish between images of cats and dogs.

TinyLlama is a lightweight 1.1B-parameter Llama model optimized for chat and programming assistance tasks.

This model is a fine-tuned image classification model based on google/vit-base-patch16-224-in21k on the CIFAR10 dataset, achieving an accuracy of 96.78%.

An X-ray image detection model fine-tuned based on AI-image-detector, achieving 99.83% accuracy

A Chinese ink painting style generation model fine-tuned based on the Stable Diffusion XL framework

This is a LoRA adapter designed for Stable Diffusion v1-5, reducing inference steps to just 2-8 steps, significantly improving generation speed.

An image classification model fine-tuned on the CIFAR-10 dataset based on the Vision Transformer (ViT) architecture, achieving 99.2% accuracy

A text-to-image generation model based on Stable Diffusion that can generate high-quality images according to text descriptions.

This is a text-to-image generation model based on Stable Diffusion, capable of generating high-quality images according to text descriptions.

BK-SDM is a stable diffusion model compressed through architecture compression, used for efficient and general text-to-image synthesis. It achieves lightweight design by removing residual and attention blocks in the U-Net.

This is an unconditional image generation model based on the diffusion model, specifically designed to generate cute butterfly images.

This is a version of thennal/whisper-medium-ml converted to the CTranslate2 model format for Malayalam speech recognition

This is a CTranslate2-optimized multilingual translation model that supports fast translation from multiple Romance languages to English.

This is a quantized version of the Helsinki-NLP/opus-mt-de-en model, enabling fast inference through CTranslate2, supporting German to English machine translation.

Lightweight image classification model based on DeiT-tiny architecture, achieving 98.56% validation accuracy after fine-tuning on a custom image dataset

A computer vision model based on ConvNeXt-Tiny architecture, fine-tuned on specific datasets for image classification tasks