# Open-source large model

Bielik-4.5B-v3.0-Instruct-GGUF is a Polish large language model released by SpeakLeash, converted from Bielik-4.5B-v3.0-Instruct to GGUF quantized format, suitable for local inference.

Qwen3-14B is a GGUF format model generated from Qwen/Qwen3-14B-Base, supporting text generation tasks and optimized for memory efficiency using IQ-DynamicGate ultra-low-bit quantization technology.

Qwen3 is the latest version of the Tongyi Qianwen series of large language models, offering a dense model with 32.8B parameters, achieving breakthrough progress in reasoning capabilities, instruction following, agent functionality, and multilingual support.

This model is a GGUF format conversion of THUDM/GLM-Z1-9B-0414, supporting Chinese and English text generation tasks.

Instella is a series of 3-billion-parameter open-source language models developed by AMD, trained on AMD Instinct™ MI300X GPUs, outperforming other fully open-source models of the same scale.

Lucie-7B is a multilingual causal language model with 7 billion parameters jointly built by LINAGORA and OpenLLM-France. It is based on the Llama-3 architecture and pre-trained on 3 trillion tokens of multilingual data.

Rombos-LLM-V2.5-Qwen-32b is a model continuously fine-tuned based on Qwen2.5-32B, with better performance than the original instruction model and base model.

A 128K long-context large language model developed based on the Llama-3 foundation model, focusing on improving RAG and long-text comprehension capabilities

A Dutch language instruction-optimized large model based on Llama-3-8B, fine-tuned through SFT and DPO on multiple Dutch instruction datasets

Llama 2 is a series of open-source large language models developed by Meta, including pre-trained and fine-tuned versions of various scales, suitable for dialogue and generation tasks.

Llama 2 is Meta's next-generation open-source large language model, available in 7B, 13B, and 70B parameter versions with commercial licensing support.

A Korean language model continuously pre-trained on the Llama-3-8B framework, trained with over 60GB of deduplicated text data

Korean text generation model based on Llama 2 architecture, specialized for Korean natural language processing tasks

JetMoE-8B is an efficient open-source large language model that achieves performance comparable to LLaMA2-7B with a training cost of under $100,000, specifically designed for low-resource environments.

The first open-source large language model to achieve an average score exceeding 80%, fine-tuned from MoMo-72B-lora-1.8.7-DPO using innovative DPO-Positive technology for preference learning optimization

Mistral-7B-Instruct-v0.2 is an instruction-finetuned large language model improved from Mistral-7B-Instruct-v0.1, compressed to 2% sparsity using Wanda pruning method while maintaining competitive performance without retraining.

ToRoLaMa is an open-source large language model for multi-turn dialogue, focusing on Vietnamese and supporting multiple international languages.

GEITje-7B is a large open-source Dutch language model based on Mistral 7B, significantly enhancing Dutch language capabilities and local knowledge through additional training on 10 billion Dutch text tokens.

Starling-7B is an open-source large language model trained through Reinforcement Learning from AI Feedback (RLAIF), fine-tuned based on Openchat 3.5, and performs excellently in MT Bench.

Mistral-7B-Instruct-v0.1 is an instruction fine-tuned version based on Mistral-7B-v0.1, suitable for dialogue generation tasks.

A 7B-parameter large language model released by Salesforce AI Research, supporting 8K long sequence input, open-sourced under Apache-2.0 license

MOSS is an open-source, plugin-extensible conversational language model developed by Fudan University. It can understand multiple languages, follow instructions in multi-round conversations, and reject inappropriate requests.