🚀 New AI Model DeepCoder-14B-Preview Released with Impressive Performance
#DeepCoder14B #AIModel #OpenSource #AIGC #CodeTesting #LiveCodeBench #Performance #TogetherAI #Agentica #MachineLearning
According to Foresight News, the AIGC development community has reported that Together AI and Agentica have jointly released an open-source model named DeepCoder-14B-Preview. This model, with 14 billion parameters, achieved a score of 60.6% on the well-known code testing platform LiveCodeBench. This performance surpasses OpenAI's o1 model, which scored 59.5%, and is slightly below the o3-mini model, which scored 60.9%. Additionally, the evaluation data on platforms such as Codeforces and AIME2024 shows that DeepCoder-14B-Preview performs nearly on par with both the o1 and o3-mini models.#DeepCoder14B #AIModel #OpenSource #AIGC #CodeTesting #LiveCodeBench #Performance #TogetherAI #Agentica #MachineLearning
🚀 Xiaomi Unveils MiMo-7B Model Surpassing Competitors in Reasoning Tasks
#Xiaomi #MiMo7B #OpenSource #Reasoning #ArtificialIntelligence #ReinforcementLearning #HuggingFace #AIME #LiveCodeBench #Algorithm
According to PANews, Xiaomi has announced the release of its first open-source model designed for reasoning, named MiMo-7B. The model has demonstrated superior performance in AIME mathematical reasoning and LiveCodeBench programming evaluations, outperforming OpenAI's o1-mini and Alibaba's Qwen-32B. MiMo-7B enhances reasoning capabilities through pre-training on large-scale reasoning data and subsequent reinforcement learning, introducing innovative algorithms such as Test Difficulty Driven Reward. It leads in reinforcement learning potential among similar 7B models. The entire series has been made available on Hugging Face, accompanied by a detailed technical report.#Xiaomi #MiMo7B #OpenSource #Reasoning #ArtificialIntelligence #ReinforcementLearning #HuggingFace #AIME #LiveCodeBench #Algorithm