AITOPAI热报
IndustryMulti-source

DeepSeek 等中国 AI 公司收入仅为 OpenAI、Anthropic 的 10%

Nikkei Asia@NikkeiAsia
Post· 11 sources
View the post
Why we picked it

日经亚洲的数据报道，DeepSeek 这些中国厂商的收入只有 OpenAI 和 Anthropic 的十分之一。如果你在对比中美模型的商业化和定价趋势，这篇值得看一眼。

据 Nikkei Asia 报道，DeepSeek 等中国 AI 公司的商业化收入只有 OpenAI 和 Anthropic 的约 10%。报道指出，尽管中国厂商在开源模型和推理成本上具备价格优势，但通过 API 和订阅服务变现的能力明显落后于美国头部实验室。

Image · Nikkei Asia
Body

收入差距的背后是产品形态与客户结构的差异：OpenAI 和 Anthropic 依托 ChatGPT、Claude 等消费级产品和企业级 API 建立了规模化付费渠道，而中国厂商更多以开源权重和低价 API 获取开发者，单用户付费转化率较低。

对开发者而言，这意味着中国模型在价格上仍将保持竞争力，而美国实验室则有更充裕的资金投入下一代模型训练。

Original · Nikkei Asia

DeepSeek, other China AI firms have just 10% the revenue of OpenAI, Anthropic https://t.co/SGSExtYRHJ

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Industry" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include DeepSeek, OpenAI, Anthropic, 行业收入.

Source attribution: curated from Nikkei Asia (@NikkeiAsia). Cross-source corroboration: 11 independent sources have covered this development (including Tech in Asia, TestingCatalog, Rest of World), confirming industry-wide signal strength.

#DeepSeek#OpenAI#Anthropic#行业收入

Keep reading

Recent stories with the same tag or section
IndustryUS vs ChinaMulti-source02:20
Chinese content中国 AI 公司挖角西方顶尖研究员，但成功率不高

Rest of World 这篇讲中国 AI 公司去美国挖研究员，结果钱给够了人家也不来。原因除了薪水差距，还有算力、研究环境和团队差距。关心中美 AI 人才竞争格局的值得看看，比单纯比模型跑分更有信息量。

#DeepSeek#人才招聘#行业观察
Event
Rest of World@restofworldSource
ProductsUS vs ChinaMulti-sourceSelected01:12
Together Compiler Hits 2.7× Throughput Gain on B200 and B300

The gap between 1.5× and 2.7× across models suggests the compiler's optimizations are architecture-dependent, so teams should benchmark their specific model before committing. Teams self-hosting or choosing an inference provider for Kimi, GLM, DeepSeek, or MiniMax models can use these figures to compare serving costs. Together's pitch targets the growing ecosystem serving Chinese open-weight models on NVIDIA's B-series hardware.

#Together AI#Kimi K3#DeepSeek#inference optimization
Event
Official accountTogether AI@togethercomputeSource
ModelsUS vs ChinaMulti-source01:03
Chinese contentFable-5.2 发布，博主对比它与 DeepSeek 的生成成本：40 美元 vs 1.4 元

博主拿刚出的 Fable-5.2 和 deepseek-v4.1-flash-1017 跑了同一个 demo，一个花 40 刀，一个只花 1 块 4，效果对比全在帖子里。做生成类应用、要算成本账的开发者可以直接看图判断值不值，这种带真实账单的横评不常见。

#Fable-5.2#DeepSeek#Anthropic#模型对比
Event
karminski-牙医 (AI工具)@karminski3Source
How-toBuildersSelected00:04
Auto-Tuner Cuts RL Training Time Up to 4.93x

The non-obvious point is that most of the speedup came from configuration, not new hardware — teams may already have equivalent capacity locked behind untuned settings. Practitioners running RL fine-tuning pipelines, especially with DeepSeek models, should benchmark their own step times against these figures before provisioning more GPUs. The 40-GPU versus 56-GPU result is a direct cost argument for adding auto-tuning to training stacks.

#reinforcement learning#auto-tuning#DeepSeek#fine-tuning
Ate-a-Pi@svpinoSource
PapersBuilders15:20
Chinese contentRAT 论文提出可动态重组的智能体框架形式化基础

做智能体框架的开发者可以看看这篇。它把 OpenHands、DeepSeek Harness 这类运行中换工具、换策略的行为用形式化语言描述清楚，还能编译成 ASP 验证。如果你的 harness 会动态挂载 MCP 工具或切换模型，这套理论能帮你检查组件变更后行为是否一致。

#RAT#智能体#Answer Set Programming#OpenHands
arXiv: DeepSeek@Jianmin JiSource
PapersBuilders13:13
Chinese contentarXiv 论文提出智能体记忆评估新协议，区分正确答案与证据充分性

做智能体记忆或评估的朋友该看看：论文用逐字节相同的可见上下文构造出答案不确定的案例，发现 Qwen3-32B 和 DeepSeek-V4-Flash 在 203 个证据不足实例里几乎不弃答、直接给参考答案。这暴露了现有记忆评测把"答对"当"答得合理"的漏洞，设计评估流程的团队尤其需要警惕。

#Qwen3#DeepSeek#智能体#评估基准
arXiv: DeepSeek@Quan Xue et al. (10)Source