# Enhanced mathematical reasoning

DeepSeek-R1-0528 is a minor version upgrade of the DeepSeek R1 model. It significantly improves the reasoning ability through increased computing resources and algorithm optimization, and performs excellently in multiple benchmark evaluations such as mathematics, programming, and general logic.

A quantized version model developed based on DeepSeek-R1-0528-Qwen3-8B, with significant improvements in reasoning ability and reduction of hallucination rate, suitable for various natural language processing tasks.

Autoregressive-32B is a Multiverse-32B baseline model built based on autoregressive modeling, providing strong support for text generation tasks.

Phi-4-reasoning-plus is the most advanced open-weight reasoning model fine-tuned by Microsoft based on Phi-4, focusing on advanced reasoning abilities in the fields of mathematics, science, and coding.

Phi-4-reasoning-plus is a large language model developed by Microsoft with enhanced reasoning capabilities, specifically optimized for complex mathematical problems and multi-step reasoning tasks.

STILL-3-1.5B-preview is a slow-thinking model enhanced with reinforcement learning technology, achieving 39.33% accuracy on the AIME benchmark

Sky-T1-32B-Preview is a 32B-parameter large language model, quantized using llama.cpp's imatrix, suitable for text generation tasks.

Carrot Llama-3.2 Rabbit Ko is a large language model fine-tuned with instructions, supporting Korean and English, and performs excellently in text generation tasks.

The Sunfall model is a natural language processing model developed based on Llama-3.1-8B-Stheno-v3.4, suitable for specific functions and application scenarios.

DeepSeek-Coder-V2 is an open-source mixture-of-experts (MoE) code language model that can achieve performance comparable to GPT4-Turbo in specific code tasks.

A model further fine-tuned based on Qwen2-7B-Instruct, excelling in handling complex multi-round tool/function calling tasks.

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

A large language model created by merging two Solar-10.7B instruction-tuned models, with performance comparable to GPT-3.5 and Gemini Pro, surpassing Mixtral-8x7b

MetaMath-Mistral-7B is a mathematical reasoning model fine-tuned on the MetaMathQA dataset based on the Mistral-7B model, which significantly improves the ability to solve mathematical problems.

WizardLM-13B V1.2 is a large language model trained on Llama-2 13b, focusing on complex instruction-following capabilities.