# Multi-task adaptation

This is a pre-trained language model merged using the mergekit tool, employing the DARE TIES method to fuse multiple Gemma3-12B variant models.

An instruction fine-tuned model optimized based on the Mistral architecture, suitable for zero-shot classification tasks

This is a 9B-parameter language model based on the Gemma-2-9B series, fused using the DARE TIES method, combining multiple optimized versions of the Gemma model.

The Gemma 3 27B Merged Model is a large language model created by linearly combining a pre-trained model and an instruction-tuned model

French text generation model fine-tuned based on canopylabs/orpheus-3b-0.1-pretrained

This is a GGUF-format converted 8B-parameter language model specialized for fantasy fiction, supporting 128K context length.

A zero-shot classifier fine-tuned based on ModernBERT-base, efficient and fast with low memory usage, suitable for various text classification tasks.

A zero-shot classifier fine-tuned based on ModernBERT-large, efficient and fast with low memory usage, suitable for various text classification tasks.

ESM++ is a faithful implementation of ESMC, supporting batch operations and fully compatible with the Huggingface ecosystem without relying on the official ESM Python package. The large version corresponds to ESMC's 600-million parameter model.

This is a 22B-parameter language model merged from Cydonia-22B-v1.3 and Magnum-v4-22B using the SLERP fusion method

This model is released under the MIT license, specific details are currently unknown.

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

BERT model for computing Russian sentence embeddings, developed based on cointegrated/LaBSE-en-ru with optimized Russian language processing performance

Open-source model based on MIT license, specific functions and uses require reference to the full documentation

A French zero-shot classification model based on Transformers.js, supporting multi-category classification of French text without specific training

MeowGPT-3.5 is a merged model based on the GPT-3.5 architecture, focusing on text generation tasks.

A pre-trained language model created using the DARE TIES merging method, based on Meta-Llama-3-8B and Meta-Llama-3-8B-Instruct models

This model is a hybrid obtained by merging the Eros-7b-test and Eris-Lelanacles-7b 7B parameter models using the SLERP method

Curie-7B-v1 is a model fine-tuned from English large language models (LLMs) for Polish text generation, excelling in Polish text generation and various NLP tasks.

This is an ultra-compact, high-performance zero-shot classification model designed specifically for edge devices and browser applications.

This is a model released under the Apache-2.0 license, with specific details currently unavailable.

This model is an open-source AI model based on the Apache-2.0 license. Specific functionalities depend on the actual model type

State-of-the-art BERT language model suite for Modern Hebrew

This model is released under the MIT license, specific details are currently unknown.

An open-source model released under the GPL-3.0 license, specific functionalities and application areas require further confirmation

This is an open-source model based on the Apache-2.0 license. Its specific functions and uses need to be determined based on the actual model.

BiT is a simple method for scaling up pre-training of ResNet-like architectures, bringing significant improvements in transfer learning.

LERT is a linguistics theory-driven pre-trained language model designed to enhance model performance through linguistic knowledge.

Open-source model based on GPL-3.0 license (specific information unavailable)

Open-source model based on the Apache-2.0 license (specific details to be added)

This is an open-source model based on the Apache-2.0 license. For specific details, please refer to the model documentation.

MiniLM v2 is a lightweight language model distilled by Microsoft from RoBERTa-Large, featuring high efficiency and compactness.

This is a TinyBERT model optimized for German, created by distilling the BERT base cased model, suitable for natural language processing tasks.

IndoBERT is an advanced language model based on BERT, specifically designed for Indonesian, trained with masked language modeling and next sentence prediction objectives.

ruT5-base is a Russian text-to-text generation model developed by SberDevices, based on the T5 architecture with 222 million parameters and trained on 300GB of data.

This is a bart-large model fine-tuned on the SQuADv2 dataset for question-answering tasks, based on the BART architecture, suitable for natural language understanding and generation tasks.

This is a tiny Russian BERT model suitable for various natural language processing tasks.