This is a large language model with 172B parameters obtained by uniformly pruning 25% of the experts in MiniMax-M2 using the REAP method. It has been optimized and quantized specifically for llama.cpp, supporting multiple quantization levels and can run in LM Studio or projects based on llama.cpp.

Natural Language ProcessingGgufEnglish

This is a large language model with 139B parameters obtained by uniformly pruning 40% of the experts in MiniMax-M2 using the REAP method. It adopts the GLM architecture and Mixture of Experts (MoE) technology, and undergoes various quantization processes through llama.cpp, suitable for text generation tasks.

Natural Language ProcessingGgufEnglish

This is the MXFP4_MOE quantized version of the MiniMax-M2-REAP-172B-A10B model, which is a memory-efficient compressed model. Through the REAP (Routing-Weighted Expert Activation Pruning) method, the model is compressed from 230B parameters to 172B parameters while maintaining performance, with a 25% reduction in size. It is suitable for resource-constrained environments, local deployment, and academic research.

Natural Language ProcessingGguf

MiniMax-M2-REAP-162B-A10B is an efficiently compressed version of MiniMax-M2. It uses the REAP (Routing Weighted Expert Activation Pruning) method to reduce the model size by 30% while maintaining almost the same performance, compressing from 230B parameters to 162B parameters, significantly reducing the memory requirement.

Natural Language ProcessingTransformersEnglish

MiniMax-M2-REAP-172B-A10B is a memory-efficient compressed variant of MiniMax-M2. It uses the REAP expert pruning method, reducing the model size by 25% from 230B parameters to 172B parameters while maintaining almost the same performance.

Natural Language ProcessingTransformersEnglish

This is the MXFP4_MOE quantized version of the MiniMax-M2-THRIFT model. It has been compressed based on the original model, including 25% expert pruning (from 256 to 192) and setting top_k=8. Meanwhile, it retains the characteristics of the encoding model and can be used for text generation tasks.

Natural Language ProcessingGguf

Qwen3-Coder-REAP-363B-A35B is a sparse mixture-of-experts model obtained by pruning 25% of the experts from Qwen3-Coder-480B-A35B-Instruct using the REAP method. While maintaining performance close to the original model, it significantly reduces the parameter scale and memory requirements, making it particularly suitable for code generation and intelligent coding scenarios with limited resources.

Natural Language ProcessingTransformersEnglish

A version of model pruning based on Qwen-Image-Edit. 20 layers are removed, and the weights of 40 layers are retained. The final model size is 13.6B parameters. This is an image editing model that supports multi-image input and text-guided image synthesis and editing.

MultimodalDiffusersMultiple Languages

This is the Q4_K_M gguf quantized version of the GLM-4.6-REAP-266B-A32B model, built on the transformers library, with text generation capabilities. The model uses the REAP (Routing-Weighted Expert Activation Pruning) method, which can selectively remove redundant experts while retaining the router's independent control over the remaining experts.

Natural Language ProcessingTransformers

This is the Q2_K_M gguf quantized version of the GLM-4.6-REAP-266B-A32B model, created based on the REAP (Router Weighted Expert Activation Pruning) method. It can selectively remove redundant experts while retaining the router's independent control over the remaining experts, achieving efficient text generation.

Natural Language ProcessingTransformers

This model is a mixture of experts model after 25% uniform pruning of GLM-4.5-Air using the REAP method. It has been converted to a 4-bit quantized version in the MLX format, suitable for efficient inference on Apple chip devices.

Natural Language ProcessingTransformersEnglish

Zirel-3 is a specific fine-tuned model based on GLM-4.5-Air-REAP-82B-A12B. It is an 82-billion-parameter mixture-of-experts model compressed using the REAP (Routing Weighted Expert Activation Pruning) technology, significantly reducing the model size while maintaining high performance.

Natural Language ProcessingTransformersEnglish

An optimized version based on the GLM-4.5-Air model with 25% expert pruning using the REAP method, converted to the MLX format for efficient operation on Apple Silicon devices

Natural Language ProcessingMlxEnglish

This is a quantized version model based on AmanPriyanshu/gpt-oss-6.0b-specialized-all-pruned-moe-only-7-experts. It adopts a mixture of experts architecture and pruning technology, providing multiple quantization types to meet the needs of different scenarios. The model supports English text generation tasks.

Natural Language ProcessingTransformersMultiple Languages

This is a 4.2B parameter Mixture of Experts model based on the GPT-OSS architecture. After pruning optimization, only 4 experts are retained, and it is converted to a quantized version in GGUF format. The model focuses on efficient inference and text generation tasks, using Q4_K_M quantization technology to reduce memory usage while maintaining performance.

Natural Language ProcessingGgufMultiple Languages

This is a quantized version of a 6-billion parameter Mixture of Experts (MoE) model based on the GPT architecture. After pruning optimization, only 7 experts are retained. It focuses on text generation tasks and is committed to making knowledge accessible to everyone.

Natural Language ProcessingGguf

This is a 6B-parameter Mixture of Experts model based on the GPT-OSS architecture. After pruning optimization, only 7 experts are retained, and it is the GGUF version using the Q4_K_M quantization format. This model focuses on efficient inference and text generation tasks, achieving better computational efficiency through pruning and quantization techniques.

Natural Language ProcessingGgufMultiple Languages

This is a specialized mixture of experts model with a scale of 6B parameters based on the GPT architecture. After pruning optimization, only 7 experts are retained, and it is converted into a quantized version in the GGUF format. The model uses Q8_0 quantization, significantly reducing memory usage and computational requirements while maintaining performance.

Natural Language ProcessingGgufMultiple Languages

A lightweight text-to-speech model optimized for Japanese, improving inference speed through layer pruning, supporting 14 voice tones

Audio ProcessingTransformersJapanese

Granite Guardian 3.2 is a streamlined risk detection model based on version 3.1, achieving more efficient inference through iterative pruning technology, focusing on identifying multi-dimensional risks in prompts and responses.

Natural Language ProcessingTransformers