# Enhanced Mathematical Reasoning

Qwen3-4B is the latest generation large language model in the Qwen series, featuring 4B parameters, supporting a 128k context window and over 100 languages, with outstanding performance in reasoning, instruction following, and agent capabilities.

Qwen3 is the latest version of the Tongyi Qianwen series of large language models, offering a complete suite of dense models and Mixture of Experts (MoE) models. Based on extensive training data, Qwen3 achieves breakthrough progress in reasoning capabilities, instruction following, agent functionalities, and multilingual support.

Qwen3-4B is the latest generation of the Qwen series' 4-billion-parameter large language model, supporting switching between reasoning and non-reasoning modes with powerful inference, instruction following, and agent capabilities.

Phi-4 Reasoning Enhanced is a 14-billion parameter open-source reasoning model developed by Microsoft Research, optimized through supervised fine-tuning and reinforcement learning, focusing on advanced reasoning capabilities in mathematics, science, and programming fields.

Qwen3 is the latest generation of large language models in the Tongyi Qianwen series, offering a complete set of dense models and Mixture of Experts (MoE) models. Based on large-scale training, Qwen3 has achieved breakthroughs in reasoning, instruction following, agent capabilities, and multilingual support.

An improved model based on Qwen2.5-Math-7B, with extended context window and optimized reasoning capabilities

A large language model trained using the DAPO algorithm based on the Qwen2.5-32B model, specializing in mathematical problem-solving and multilingual text generation

ReasonFlux-F1-32B is a hierarchical large language model based on template-enhanced reasoning trajectory fine-tuning, excelling in reasoning tasks.

A small-scale large language model enhanced by reinforcement learning, focused on improving the reasoning capabilities of a 1.5B parameter model

A large language model optimized based on Qwen's QwQ 32B architecture, focusing on efficient memory utilization, programming assistance, and complex problem-solving.

This is a 14B-parameter large language model merged using the SLERP method, combining two base models: li-14b-v0.4 and miscii-14b-0218.

Phi-4-Model-Stock-v2 is a large language model merged from multiple Phi-4 variant models using the model_stock merging method, demonstrating strong performance across multiple benchmarks.

LuxLlama is a fine-tuned version of the Meta-Llama-3.1-8B-Instruct model, specifically optimized for Luxembourgish language understanding and generation, as well as general and mathematical reasoning capabilities.

Skywork o1 Open-PRM-Qwen-2.5-1.5B is an incremental process reward model trained on Qwen2.5-Math-1.5B-Instruct, specifically designed to enhance small-scale complex problem-solving capabilities.

A 7B-parameter model in the Skywork o1 open model series, trained on Qwen2.5-Math-7B-Instruct, featuring enhanced reasoning capabilities through progressive process rewards

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.

A Mixture of Experts large language model based on 3 Mistral architecture models, excelling in dialogue, code, and mathematical tasks

Based on the Mistral-7b model, it employs the Quiet-STaR method for continuous pretraining, generating 8 reasoning tokens before each output token to enhance reasoning capabilities.

A large language model fine-tuned based on jondurbin/bagel-34b-v0.2, optimized with novel DPO-Positive (DPOP) technology for preference learning

OpenChat is an open-source language model based on Mistral-7B, trained with mixed-quality data to deliver exceptional text generation capabilities.

OpenChat 3.5 1210 is an open-source language model based on the Mistral-7B architecture, trained on mixed-quality datasets, supporting programming, general dialogue, and mathematical reasoning tasks, with overall performance surpassing ChatGPT (March version) and Grok-1.

Experimental 8x7b Mixture of Experts model developed based on Mistral AI's Mixtral 8x7b, fine-tuned on Synthia, MetaMathQA, and Capybara datasets

MetaMath-Llemma-7B is a math-specialized large language model fully fine-tuned on the MetaMathQA dataset based on the Llemma-7B foundation model, significantly improving mathematical problem-solving capabilities.

MetaMath-Llemma-7B is a mathematical reasoning model fine-tuned on the MetaMathQA dataset, demonstrating excellent performance on GSM8K and MATH datasets.

WizardMath is a large language model empowered by Reinforced Learning from Evolved Instruction Feedback (RLEIF), specifically designed to enhance mathematical reasoning capabilities.