（2025·南京市外国语学校期中改编）DeepSeek - R1, a new AI model developed by the Hangzhou - based Chinese company DeepSeek, has rocked the global AI industry. Released on Jan. 20, it quickly beat OpenAI’s ChatGPT and topped the free charts of Apple’s App Store by Jan. 27.
According to DeepSeek, in tasks like maths and natural language reasoning（推理）, the performance of this model matches that of the leading models from big companies like OpenAI. But it’s much cheaper and uses much less computing power（算力）. The report on its earlier V3 model showed that DeepSeek is the least expensive among large language models（LLMs）. It costs about 5.57 million US dollars（about 40.58 million yuan）to make it.
Behind the lower cost is a major innovation in R1’s training. Different from traditional methods such as Chain - of - Thought（CoT）and Supervised Fine - Tuning（SFT）, DeepSeek uses Reinforcement Learning（RL）as the main training method. While CoT depends on step - by - step reasoning and SFT on huge quantities of data, RL enables models to learn through interaction and reward systems. This has allowed R1 to develop reasoning skills in a natural way, making it better suited for complex and changing tasks.
The AI industry development has long depended on making computing power bigger. “By significantly increasing the data quality and improving the model architecture, DeepSeek may transform the AI field,” pointed out the US investment bank Morgan Stanley. “Bigger is no longer always smarter,” noted the bank.
What’s more, DeepSeek - R1 is open - source. Everyone is free to access the model’s code（代码）and make changes to suit their own needs. “We won’t choose closed - source,” said Liang Wenfeng, the founder of DeepSeek. He highlighted the importance of sharing knowledge, building a healthy ecosystem and contributing to the overall progress of technology. “That is the power of open research and open source,” Meta’s chief AI scientist Yann LeCun agreed.
1. （新考法 组合选择）What are the advantages of DeepSeek compared with other similar technologies?
① Costing less to make.
② Being free to users.
③ Being open - source.
④ Having bigger computing power.
A. ①③④
B. ①②④
C. ①②③
D. ②③④
2. （新考法 句意理解）What does the underlined sentence in Para. 4 mean?
A. The bigger investment is, the smarter the AI model is.
B. Computing power is still an important factor in developing AI models.
C. AI models with the biggest computing power are not necessarily the smartest.
D. AI models with less computing power are good at difficult tasks as other big models.
3. What is the best title for the passage?
A. DeepSeek: Future king in the AI field
B. DeepSeek: Learning to think on its own
C. DeepSeek: Useful tool to help people make choices
D. DeepSeek: Changing the game rules with lower costs
According to DeepSeek, in tasks like maths and natural language reasoning（推理）, the performance of this model matches that of the leading models from big companies like OpenAI. But it’s much cheaper and uses much less computing power（算力）. The report on its earlier V3 model showed that DeepSeek is the least expensive among large language models（LLMs）. It costs about 5.57 million US dollars（about 40.58 million yuan）to make it.
Behind the lower cost is a major innovation in R1’s training. Different from traditional methods such as Chain - of - Thought（CoT）and Supervised Fine - Tuning（SFT）, DeepSeek uses Reinforcement Learning（RL）as the main training method. While CoT depends on step - by - step reasoning and SFT on huge quantities of data, RL enables models to learn through interaction and reward systems. This has allowed R1 to develop reasoning skills in a natural way, making it better suited for complex and changing tasks.
The AI industry development has long depended on making computing power bigger. “By significantly increasing the data quality and improving the model architecture, DeepSeek may transform the AI field,” pointed out the US investment bank Morgan Stanley. “Bigger is no longer always smarter,” noted the bank.
What’s more, DeepSeek - R1 is open - source. Everyone is free to access the model’s code（代码）and make changes to suit their own needs. “We won’t choose closed - source,” said Liang Wenfeng, the founder of DeepSeek. He highlighted the importance of sharing knowledge, building a healthy ecosystem and contributing to the overall progress of technology. “That is the power of open research and open source,” Meta’s chief AI scientist Yann LeCun agreed.
1. （新考法 组合选择）What are the advantages of DeepSeek compared with other similar technologies?
① Costing less to make.
② Being free to users.
③ Being open - source.
④ Having bigger computing power.
A. ①③④
B. ①②④
C. ①②③
D. ②③④
2. （新考法 句意理解）What does the underlined sentence in Para. 4 mean?
A. The bigger investment is, the smarter the AI model is.
B. Computing power is still an important factor in developing AI models.
C. AI models with the biggest computing power are not necessarily the smartest.
D. AI models with less computing power are good at difficult tasks as other big models.
3. What is the best title for the passage?
A. DeepSeek: Future king in the AI field
B. DeepSeek: Learning to think on its own
C. DeepSeek: Useful tool to help people make choices
D. DeepSeek: Changing the game rules with lower costs
答案：一 1. C
2. C 【解析】根据第四段内容可知，不同于以往的人工智能产业依赖于提高计算能力，DeepSeek 通过提高数据质量和模型架构改进性能，由此可推知，画线句子指计算能力最强的模型不一定是最智能的。
3. D
2. C 【解析】根据第四段内容可知，不同于以往的人工智能产业依赖于提高计算能力，DeepSeek 通过提高数据质量和模型架构改进性能，由此可推知，画线句子指计算能力最强的模型不一定是最智能的。
3. D
解析：
翻译:
（2025·南京外国语学校期中改编）总部位于杭州的中国公司深度求索（DeepSeek）研发的新型人工智能模型DeepSeek - R1，震撼了全球人工智能行业。该模型于1月20日发布，到1月27日就迅速超越了OpenAI的ChatGPT，登上了苹果应用商店的免费排行榜榜首。
据深度求索介绍，在数学、自然语言推理等任务中，该模型的性能可与OpenAI等大公司的领先模型相媲美。但它的成本要低得多，所需的算力也少得多。关于其早期版本V3模型的报告显示，深度求索在大型语言模型中成本最低，研发它大约花费了557万美元（约合4058万元人民币）。
成本降低的背后，是R1在训练方面的一项重大创新。与思维链（CoT）、有监督微调（SFT）等传统方法不同，深度求索将强化学习（RL）作为主要训练方法。思维链依赖于逐步推理，有监督微调依赖于大量数据，而强化学习能让模型通过交互和奖励机制来学习。这使得R1能够以一种自然的方式发展推理能力，从而更适合复杂多变的任务。
长期以来，人工智能行业的发展都依赖于提升算力。美国投资银行摩根士丹利指出：“通过显著提高数据质量和改进模型架构，深度求索可能会改变人工智能领域的格局。”该银行还指出：“更大（算力）并不总是更聪明。”
此外，DeepSeek - R1是开源的。任何人都可以自由访问该模型的代码，并根据自身需求进行修改。深度求索的创始人梁文峰表示：“我们不会选择闭源。”他强调了知识共享、构建健康生态系统以及推动技术整体进步的重要性。“这就是开放研究和开源的力量。”Meta的首席人工智能科学家杨立昆也表示认同。
1. （新考法 组合选择）与其他同类技术相比，深度求索有哪些优势？
① 研发成本更低。
② 对用户免费。
③ 是开源的。
④ 算力更强。
A. ①③④
B. ①②④
C. ①②③
D. ②③④
2. （新考法 句意理解）第4段中划线句子是什么意思？
A. 投资越大，人工智能模型就越智能。
B. 算力仍然是开发人工智能模型的一个重要因素。
C. 算力最强的人工智能模型不一定是最智能的。
D. 算力较弱的人工智能模型和其他大型模型一样擅长处理困难任务。
3. 这篇文章的最佳标题是什么？
A. 深度求索：人工智能领域未来的王者
B. 深度求索：学会自主思考
C. 深度求索：帮助人们做选择的实用工具
D. 深度求索：以更低成本改变游戏规则
（2025·南京外国语学校期中改编）总部位于杭州的中国公司深度求索（DeepSeek）研发的新型人工智能模型DeepSeek - R1，震撼了全球人工智能行业。该模型于1月20日发布，到1月27日就迅速超越了OpenAI的ChatGPT，登上了苹果应用商店的免费排行榜榜首。
据深度求索介绍，在数学、自然语言推理等任务中，该模型的性能可与OpenAI等大公司的领先模型相媲美。但它的成本要低得多，所需的算力也少得多。关于其早期版本V3模型的报告显示，深度求索在大型语言模型中成本最低，研发它大约花费了557万美元（约合4058万元人民币）。
成本降低的背后，是R1在训练方面的一项重大创新。与思维链（CoT）、有监督微调（SFT）等传统方法不同，深度求索将强化学习（RL）作为主要训练方法。思维链依赖于逐步推理，有监督微调依赖于大量数据，而强化学习能让模型通过交互和奖励机制来学习。这使得R1能够以一种自然的方式发展推理能力，从而更适合复杂多变的任务。
长期以来，人工智能行业的发展都依赖于提升算力。美国投资银行摩根士丹利指出：“通过显著提高数据质量和改进模型架构，深度求索可能会改变人工智能领域的格局。”该银行还指出：“更大（算力）并不总是更聪明。”
此外，DeepSeek - R1是开源的。任何人都可以自由访问该模型的代码，并根据自身需求进行修改。深度求索的创始人梁文峰表示：“我们不会选择闭源。”他强调了知识共享、构建健康生态系统以及推动技术整体进步的重要性。“这就是开放研究和开源的力量。”Meta的首席人工智能科学家杨立昆也表示认同。
1. （新考法 组合选择）与其他同类技术相比，深度求索有哪些优势？
① 研发成本更低。
② 对用户免费。
③ 是开源的。
④ 算力更强。
A. ①③④
B. ①②④
C. ①②③
D. ②③④
2. （新考法 句意理解）第4段中划线句子是什么意思？
A. 投资越大，人工智能模型就越智能。
B. 算力仍然是开发人工智能模型的一个重要因素。
C. 算力最强的人工智能模型不一定是最智能的。
D. 算力较弱的人工智能模型和其他大型模型一样擅长处理困难任务。
3. 这篇文章的最佳标题是什么？
A. 深度求索：人工智能领域未来的王者
B. 深度求索：学会自主思考
C. 深度求索：帮助人们做选择的实用工具
D. 深度求索：以更低成本改变游戏规则