# Efficient Text Generation

SimpleStories is a micro model family designed for interpretability research, trained on the SimpleStories dataset, focusing on story generation tasks.

GLM-4-9B-0414-abliterated is a large language model with 9B parameters based on the GLM architecture, suitable for text generation tasks.

This model is a GGUF-format conversion of canopylabs/orpheus-3b-0.1-ft, suitable for text generation tasks.

GGUF quantized version of the Gemma 3 27B instruction-tuned model, suitable for text generation tasks.

This is a quantized version of a large language model, aiming to make knowledge more accessible and usable.

DeepSeek-R1-Distill-Llama-8B is an 8B-parameter large language model based on the Llama architecture, optimized through distillation training for text generation tasks.

A 200M-parameter model based on the Qwen architecture, pretrained from scratch using the MiniPLM knowledge distillation framework

Hymba-1.5B-Base is a foundational text generation model developed by NVIDIA, employing a hybrid architecture combining Mamba and attention heads, suitable for various natural language generation tasks.

This is a multilingual image-to-text model customized for internal use cases, supporting both English and German.

PowerLM-3B is a small language model with 3 billion parameters, trained using the Power learning rate scheduler, and demonstrates excellent performance across multiple benchmarks including natural language multiple-choice, code generation, and mathematical reasoning.

An improved model based on Llama2 7B, supporting hierarchical skip and self-speculative decoding to enhance inference efficiency

GGUF quantized version of Mistral-7B-Instruct-v0.3, suitable for local inference text generation models

Mistral-7B-v0.3 is an upgraded large language model based on Mistral-7B-v0.2, with the main improvement being the expansion of the vocabulary to 32,768 tokens.

Mistral-7B-Instruct-v0.3 is an instruction-tuned version of Mistral-7B, focusing on text generation tasks that follow human instructions.

Mamba-7B is a 7-billion-parameter model based on the Mamba architecture, trained over multiple rounds on the RefinedWeb dataset (1.2 trillion tokens). Mamba is a state space model that does not use self-attention mechanisms and excels in various natural language benchmarks.

Mamba is a transformer-compatible sequence modeling model with efficient inference capabilities.

Mamba is an efficient language model based on the State Space Model (SSM) architecture, with 1.4B parameters, supporting text generation tasks

DareBeagle-7B is a 7B-parameter large language model obtained by merging mlabonne/NeuralBeagle14-7B and mlabonne/NeuralDaredevil-7B using LazyMergekit, demonstrating excellent performance across multiple benchmarks.

DeciLM-7B is a pure decoder-based text generation model with 7.04 billion parameters, released under the Apache 2.0 license. It is currently the best-performing 7B base language model on the Open LLM Leaderboard.