# Large language model

The 4-bit quantization version of Tencent Hunyuan A13B large language model, suitable for instruction following tasks

A quantized version of Kimi-Dev-72B, using advanced nonlinear optimal quantization and multi-head latent attention mechanism to reduce storage and computing requirements.

A quantized version of the Austral-24B-Winton model of Delta-Vector, quantized using the llama.cpp tool, suitable for efficient operation on different hardware configurations.

A 6-bit quantized version converted from the DeepSeek-R1-0528-Qwen3-8B model, suitable for text generation tasks in the MLX framework.

This is a 4-bit quantized version converted from the Qwen3-235B-A22B-8bit model, optimized for the MLX framework and suitable for text generation tasks.

DeepSeek-R1-0528 is a large language model that has been quantized to optimize its running efficiency on different hardware.

This is the MLX format version of the THUDM/GLM-4-32B-0414 model, processed with 4-bit DWQ quantization, suitable for efficient inference on Apple silicon devices.

A 4-bit quantized version of the Gemma 3 12B model, suitable for the MLX framework and supporting efficient text generation tasks.

This is a 4-bit quantized model converted from Qwen/Qwen3-30B-A3B to MLX format, suitable for text generation tasks.

Qwen3-30B-A3B-4bit-DWQ-0508 is a 4-bit quantized model converted from Qwen/Qwen3-30B-A3B to MLX format, suitable for text generation tasks.

An open-source code reasoning large language model developed by NVIDIA, with a parameter scale of 14 billion, focused on code generation and reasoning tasks.

Qwen3-14B-4bit-AWQ is an MLX-format model converted from Qwen/Qwen3-14B, using AWQ quantization technology to compress the model to 4bit, suitable for efficient inference on the MLX framework.

Russian-optimized large language model based on Qwen3-8B, specifically designed for Russian text generation tasks

Qwen3-8B-4bit-AWQ is a 4-bit AWQ quantized version converted from Qwen/Qwen3-8B, suitable for text generation tasks in the MLX framework.

The GGUF quantized version of Qwen3-30B-A3B, supporting multi-bit quantization, suitable for text generation tasks.

Qwen3-14B-4bit is a 4-bit quantized version of the Qwen/Qwen3-14B model converted using mlx-lm, suitable for text generation tasks.

This model is a 4-bit quantized version converted from THUDM/GLM-Z1-32B-0414, suitable for text generation tasks.

This is a merged model based on Qwen2.5-32B, incorporating features from DeepSeek-R1-Distill-Qwen-32B, QwQ-32B, and Sky-T1-32B-Flash to enhance performance.

INT4 weight-quantized version of DeepSeek-R1, reducing GPU memory and disk space requirements by approximately 50% while maintaining original model performance.

Gemma 3 27B IT QAT BF16 is a version of the Gemma series of models released by Google. It has undergone quantization-aware training (QAT) and is converted to the BF16 format, suitable for the MLX framework.

Gemma 3 27B IT QAT 8bit is an MLX-format model converted from Google's Gemma 3 27B model, supporting image-to-text tasks.

The first open-source, native 1-bit large language model developed by Microsoft Research, with a parameter scale of 2 billion, trained on a corpus of 4 trillion tokens.

An open-source native 1-bit large language model developed by Microsoft Research, with 2 billion parameters trained on a 4 trillion token corpus, significantly improving computational efficiency.

PLaMo-Embedding-1B is a Japanese text embedding model developed by Preferred Networks, demonstrating outstanding performance in Japanese text embedding benchmarks

Cantonese LLM Chat v1.0 is the first-generation Cantonese large language model developed by the hon9kon9ize team, excelling in Hong Kong-related professional knowledge and Cantonese conversation.

FP8 quantized version of QwQ-32B, reducing storage and memory requirements by 50% through dynamic quantization while maintaining 99.75% of the original model accuracy

The GGUF quantization version of PLLuM-8x7B-chat, optimized for local inference, supporting multiple quantization levels to meet different hardware requirements.

Qwenvergence-14B-v13-Prose-DS is a large language model based on the merger of multiple models, with excellent data performance and humorous expression ability.

The Jais series is a bilingual large language model specialized in Arabic language processing, with strong English capabilities and 1.3 billion parameters.

Jais series 1.3 billion parameter Arabic-English bilingual large language model, optimized for exceptional Arabic capabilities while maintaining strong English proficiency

This is a GGUF format model converted from Google's Gemma 2B model, suitable for text generation tasks.

INT4 auto-quantized model based on Qwen2-7B, generated by Intel's auto-round tool, suitable for efficient inference tasks

Norwegian language adapter developed based on Llama-3-70b-fp16, trained on 1 million Norwegian text tokens

A fine-tuned version based on the EleutherAI_pythia-1b-deduped model for generating concise summaries

A Korean language model continuously pre-trained based on Llama-3-8B, using over 60GB of deduplicated publicly available text for training, supporting Korean and English text generation.

adeBERT is a fine-tuned model based on the BERT-large architecture, focusing on domain-specific tasks and demonstrating excellent performance on evaluation datasets.

This is a 4-bit quantized GPTQ model based on Meta Llama 3, quantized by Astronomer, capable of efficient operation on low-VRAM devices.

SambaLingo-Arabic-Chat-70B is a human-aligned conversational model supporting Arabic and English, adapted and trained based on Llama-2-70b.

A Russian text summarization model developed by SberDevices, based on the T5 architecture with 1.7B parameters

A 103B parameter hybrid model based on the leaked Miqu model, supporting 32K context length