# Long Context Support

MedGemma is a medical multimodal model optimized based on Gemma 3, specifically designed for medical text and image understanding, available in 4B and 27B versions.

Qwen3-0.6B is the latest 0.6B-parameter large language model in the Qwen series, supporting mode switching between reasoning and non-reasoning modes, with powerful reasoning, instruction-following, and multilingual capabilities.

A role-play specialized large language model fine-tuned from Qwen3-30B-A3B-NSFW-JP, supporting long Japanese text generation

Superthoughts Lite v2 is a lightweight Mixture of Experts (MOE) model based on the Llama-3.2 architecture, focusing on reasoning tasks to provide higher accuracy and performance.

The latest version of the Tongyi Qianwen series of large language models, supporting switching between thinking and non-thinking modes, with powerful reasoning, multilingual, and agent capabilities.

A roleplay fine-tuned version based on GLM-4-32B-0414, with excellent performance, distinctive personality, diverse styles, and elegant writing.

Qwen3-1.7B is the latest generation of the Qwen series with 1.7B parameters, supporting switching between thinking and non-thinking modes, featuring enhanced reasoning capabilities and multilingual support.

Qwen3-0.6B is a 0.6B-parameter large language model developed by Alibaba Cloud, the latest member of the Qwen3 series, supporting over 100 languages with strong reasoning, instruction-following, and multilingual capabilities.

Viper-Coder-v1.7-Vsm6 is a large language model based on the Qwen2.5 14B modal architecture, focusing on improving coding efficiency and computational reasoning capabilities, optimizing memory usage, and reducing redundant text generation.

A merged model based on Llama-3-70B, specializing in story generation and role-play (RP) tasks, combining multiple high-quality models via the breadcrumbs_ties method

The EXAONE Deep series models excel in reasoning tasks such as mathematics and programming. The 7.8B version outperforms open-source models of similar scale and even surpasses certain proprietary models.

Turkish pre-trained model based on ModernBERT architecture, supporting 8192 context length with excellent performance across multiple domains

Ola-7B is a multimodal large language model jointly developed by Tencent, Tsinghua University, and Nanyang Technological University. Based on the Qwen2.5 architecture, it supports processing text, image, video, and audio inputs and generates text outputs.

Falcon3 7B-IT and 7B-IT Mixture of Experts model with 13.4 billion parameters, supporting English, French, Spanish, and Portuguese, with a context length of up to 32K.

Jina Embeddings V2 Base Code is a transformer-based English text embedding model, specializing in feature extraction and sentence similarity calculation for code-related texts.

Slush is a two-stage model trained with high LoRA dropout rate, focusing on enhancing creativity and role-playing capabilities

A large language model with 72B parameters fine - tuned based on Qwen2.5 - 72B - Instruct. It uses dynamic FP8 quantization technology to optimize inference efficiency and aims to reproduce the prose quality of Claude 3.

Allegro is an open-source high-quality text-to-video generation model capable of producing 6-second detailed videos at 720x1280 resolution and 15 FPS.

A large language model fine-tuned on the BeyondDialogue dataset, specifically designed for Chinese-English role-playing scenarios

InternVideo2-Chat-8B-InternLM2.5 is a video-text multimodal model that enhances video understanding and human-computer interaction by integrating the InternVideo2 video encoder with a large language model (LLM).

DeepSeek-V2-Lite is a cost-efficient Mixture of Experts (MoE) language model with a total of 16B parameters and 2.4B active parameters, supporting a 32k context length.

A German-optimized large language model based on Meta Llama3-8B, continuously pre-trained on 65 billion German tokens, specifically optimized for German and supporting 32k long context

Erosumika-7B-v3 is a 7.1bpw exl2 quantized language model suitable for running 16k context on GPUs with 8GB VRAM. It was created by fusing multiple models using the DARE TIES method, primarily for entertainment-oriented fictional writing.

The first large-scale Greek foundational language model, based on the Mistral-7B architecture, enhanced with 40 billion tokens of Greek and English corpus to improve Greek language capabilities

A 70B-parameter large language model merged using the DARE linear fusion method, optimized for role-playing and story creation

Code Llama is a series of code generation and understanding models released by Meta, ranging from 7 billion to 70 billion parameters. This model is the 70 billion parameter instruction fine-tuned version.

Midnight-Miqu-70B-v1.5 is a 70B-parameter large language model specifically designed for role-playing and story creation, merged from models by sophosympatheia and migtissera.

A 7B-parameter large language model focused on role-playing, featuring 8192 context length with emphasis on creativity and reduced clichés

Code Ninja is an enhanced version of the renowned model openchat/openchat-3.5-1210, trained through supervised fine-tuning on two large-scale datasets containing over 400,000 coding instructions.

A language model scaled down from Mistral 7B to 248 million parameters, designed for text generation tasks and suitable for downstream task fine-tuning.

CodeS-7B is a large language model optimized for SQL generation, based on incremental pre-training of StarCoderBase-7B, supporting a maximum context length of 8,192 tokens.

A multilingual instruction-following language model based on LLaMA 7B, supporting English, Chinese, and Japanese, suitable for chatbots and instruction-following tasks.