# Chain-of-thought reasoning

Stockmark-2-VL-100B-beta is a Japanese-specific vision-language model with 100 billion parameters, equipped with chain-of-thought (CoT) reasoning ability and can be used for document reading and comprehension.

A large language model fine-tuned based on Google's Gemma-3 27B, specializing in chain-of-thought reasoning and code generation tasks, with optimized GPU memory and system memory management.

An instruction-tuned model based on YandexGPT-5-Lite-8B-pretrain, fine-tuned with Russian datasets GrandMaster-PRO-MAX and Grounded-RAG-RU-v2, excelling in Russian and English tasks.

AceMath is a series of cutting-edge models designed specifically for mathematical reasoning. It is improved based on Qwen and is good at using chain-of-thought (CoT) reasoning to solve English mathematical problems.

A multi-functional large language model based on the Mixtral architecture, integrating deep mind and self-optimizing capabilities, supporting multi-language text generation and complex reasoning tasks

An 8B parameter model built on Meta's Llama-3, fine-tuned using the Bagel dataset, and adopting a single llama-3 instruction format.

Minotaur 13B is an instruction fine-tuned model based on LlaMA-13B, using entirely open-source datasets for fine-tuning to ensure reproducibility.