本周新模型
12
↗ 近 14d
OpenRouter
API 均价 (输入)
$0.80
$/M tokens · 中位数
AI 股指日均涨幅
-4.08%
20 支 AI 股
新闻活跃度
136
↗ 7d
英 76 / 中 60
| # | 模型 / 厂商 | 综合分 | 趋势 | $/M (in/out) | 性价比 |
|---|---|---|---|---|---|
| 1 | DeepSeek V4.1 Flash 1017 · DeepSeek | 1427 | ▲ 12 | $0.15 / $0.60 | 100 |
| 2 | Glimpse · Elorian | 1422 | ▲ 6 | $0.60 / $3.00 | 100 |
| 3 | Beam 1.0 501B A23B · Reflection | 1417 | ▲ 4 | $1.00 / $3.00 | 100 |
| 4 | Moondream 3 Preview · Moondream | 1356 | — 0 | $0.30 / $2.50 | 58 |
| 5 | Moondream 3.1 · Moondream | 1348 | — 0 | $0.30 / $2.50 | 58 |
| 6 | Muse Spark Open · Meta | 1402 | — 0 | $1.25 / $4.25 | 90 |
| 7 | Nemotron 3.5 Super VL · NVIDIA | 1397 | — 0 | $0.10 / $0.40 | 100 |
| 8 | Pareto 26.10 · Unbiased | 1392 | — 0 | $0.80 / $3.20 | 100 |
| 9 | Pareto Latest · ~unbiased | 1387 | — 0 | $0.80 / $3.20 | 100 |
| 10 | Pareto 26.9 · Unbiased | 1382 | — 0 | $2.50 / $7.50 | 70 |
| 11 | LFM2.5-24B-A2B · Liquid | 1300 | — 0 | $0.05 / $0.70 | 82 |
| 12 | Gemini 4 Argon · Google | 1372 | — 0 | $2.00 / $10 | 75 |
DeepSeek V4.1 Flash 1017
2026-10-17
DeepSeek
DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash
Glimpse
2026-10-17
Elorian
Glimpse is a multimodal visual reasoning model from Elorian, built for visual un
Beam 1.0 501B A23B
2026-10-16
Reflection
Beam 1.0 is Reflection's open-weight reasoning model for coding and long-horizon
Moondream 3 Preview
2026-10-15
Moondream
Moondream 3 Preview is a vision-language model from Moondream with 9B total para
Moondream 3.1
2026-10-15
Moondream
Moondream 3.1 is a vision-language model from Moondream, the successor to Moondr
Muse Spark Open
2026-10-15
Meta
Muse Spark Open is a coding-optimized open-weight model. Trained specifically fo
Nemotron 3.5 Super VL
2026-10-15
NVIDIA
NVIDIA Nemotron 3.5 Super VL is an open multimodal mixture-of-experts model from
Pareto 26.10
2026-10-15
Unbiased
Pareto is a multimodal composite model built for research, coding, and agentic w
Pareto Latest
2026-10-13
~unbiased
This model always redirects to the latest model in the Pareto family.
Pareto 26.9
2026-10-13
Unbiased
Pareto is a multimodal composite model built for research, coding, and agentic w
LFM2.5-24B-A2B
2026-10-13
Liquid
LFM2.5-24B-A2B is a mixture-of-experts reasoning model from Liquid AI with 24B t
Gemini 4 Argon
2026-10-13
Google
Gemini 4 Argon is Google's flagship reasoning model, built for complex software
📱 C 端应用热度
公开统计 · 季度更新
ChatGPT
OpenAI
4.2 亿
+5.8%
Claude
Anthropic
1.1 亿
+18.4%
Gemini
Google
8200 万
+9.2%
Copilot
Microsoft
7500 万
+6.5%
Perplexity
Perplexity
4500 万
+22.0%
Grok
xAI
3000 万
+35.0%
豆包
字节跳动
7800 万
+12.1%
文心一言
百度
5200 万
+4.3%
Kimi
月之暗面
4800 万
+19.5%
通义千问
阿里
4100 万
+8.2%
智谱清言
智谱
1600 万
+11.7%
DeepSeek Chat
DeepSeek
3500 万
+45.0%
Midjourney
Midjourney
1800 万
+3.1%
Character.AI
Character.AI
2300 万
-8.4%
单卡按需时租 · 每小时美金(H100 8 卡训练 1 天 ≈ 价格 × 192)
B200
Blackwell 旗舰
暂无
H200
最新训练卡
暂无
H100 80G
训练主力
暂无
A100 80G
上一代训练
暂无
L40S
推理优化
暂无
A40
推理通用
暂无
A30
低功耗推理
暂无
L4
边缘推理
$0.37/h
T4
入门推理
暂无
RTX 6000 Ada
工作站旗舰
暂无
RTX A6000
Ampere 工作站
暂无
RTX A5000
工作站中端
暂无
V100
老一代训练
$0.14/h
RTX 5090
消费旗舰
暂无
RTX 5080
次旗舰
$0.24/h
RTX 4090
性价比之王
暂无
RTX 4080
中高端
$0.24/h
RTX 3090
上代消费
$0.24/h
RTX 3080
上代中端
$0.18/h
RTX 2080 Ti
老一代消费
$0.11/h
📈 AI 概念股
新浪财经 · 实时
2026-10-18 10:15
NVDA
英伟达 · GPU
$214.13
-1.43%
MSFT
微软 · Azure/OpenAI
$529.05
-1.93%
GOOGL
谷歌 · Gemini
$340.59
-2.47%
META
Meta · Llama
$709.33
-3.72%
AMZN
亚马逊 · AWS/Bedrock
$250.00
-1.87%
AAPL
苹果 · 端侧 AI
$343.46
+0.43%
TSLA
特斯拉 · FSD/Optimus
$354.72
-2.13%
AMD
AMD · MI 系列
$528.75
-5.95%
AVGO
博通 · 定制 AI 芯片
$328.75
-4.52%
TSM
台积电 · 代工
$412.61
-2.28%
ARM
ARM · 架构授权
$227.80
-6.94%
MU
美光 · HBM 内存
$878.84
-5.63%
QCOM
高通 · 端侧推理
$172.11
-3.53%
SMCI
超微 · AI 服务器
$31.63
-9.86%
ORCL
甲骨文 · AI 云
$128.23
-4.64%
CRM
Salesforce · Agent
$240.70
-2.45%
NOW
ServiceNow · 企业 AI
$141.97
-4.44%
PLTR
Palantir · AI 平台
$186.03
-7.22%
SNOW
雪花 · 数据 AI
$306.84
-8.43%
ADBE
Adobe · 创作 AI
$245.20
-2.57%
💰 融资动态
从新闻流自动提取
🔗 实时
36氪首发 | 大疆前高管押注个人出行机器人,累计融资近亿美元
—
资讯 · 作者 | 张子怡 编辑 | 袁斯来 硬氪获悉,AI智能硬件与机器人企业「若创科技」(英文名:Strutt)近日完成Pre
中信证券:传统主业稳健+AI收入渐显,美股应用软件迎来估值修复契机
—
资讯 · 36氪获悉,中信证券研报表示,当下“AI吞噬软件”叙事正在逐步证伪:龙头厂商收入与在手订单仍保持稳健增长,应用软件基本面
“哲源科技”完成近2亿元A2轮融资
2亿
资讯 · 36氪获悉,北京哲源科技有限责任公司在A2轮获得近2亿人民币融资。本轮融资由圣湘生物携产业基金领投,中科创星作为老股东持
📰 国际 AI 新闻
HuggingFace · VentureBeat · HN
2026-10-18 10:15
热议
90% of US working hours are spent on tasks not covered well by AI benchmarks
Hacker News · 598 分 · 542 评论
热议
Banks are now trying to offload the AI data center debt they just underwrote
Hacker News · 384 分 · 277 评论
热议
Containment and monitoring failures are to blame for AI hacking reports: experts
Hacker News · 948 分 · 620 评论
热议
I posted to HN, and then my website was flooded with AI bot traffic
Hacker News · 351 分 · 186 评论
热议
Character.ai has initiated voluntary Chapter 11 bankruptcy
Hacker News · 334 分 · 294 评论 · https://www.theverge.com/ai-artificial-intelligence/1009802/... https://techcrunch.com/2026/10/13/just-5-years-in-interactiv... https://
热议
AI is a tool and using it well takes skill
Hacker News · 802 分 · 607 评论
查看全部 76 条 →
📰 中文 AI 新闻
36氪
资讯
马斯克称硬件制造除中国外无对手,将用火箭技术造数据中心
在SpaceX公布二季度财报的电话会上,SpaceX创始人马斯克在回答提问时表示,SpaceX和特斯拉是地球上硬件实力最强的两家公司。“除了中国之外,也许算上中国也是”。 马斯克表示,会将从火箭和卫星领域积累的专业知识应用于地面数据中心的规模化建设,“我们发现,即使将制造火箭所积累的少量经验应用到数据中心建设中,也能带来巨大的收益。” 马斯克还表示,预计未来的数据中心在电力供应和冷却能力需求将远超预期。他同时透露,已经和英伟达讨论,预计明年将获得其相当一部分GPU产能。(新浪科技)
资讯
早期项目 | 从智能喂养切入,一家创业公司想搭建母婴AI生态
作者 | 张子怡 编辑 | 袁斯来 母婴市场在全球拥有万亿级规模。中国的母婴消费市场规模已突破5万亿元,年均增速保持12%的稳健水平;全球母婴市场总量更是达到约2万亿美元。 庞大的母婴市场背后,其智能化渗透率较低,尚不足1%。全球母婴家庭普遍面临养育压力大、成本高、信息繁杂等核心痛点,而传统母婴产品仍停留在基础功能阶段,智能化程度不足、生态割裂,无法满足新一代家庭需求。以喂奶为例,不少家庭依然采用着极为原始的“手背测奶温”、手工记录喂养数据等方式。 在全球超万亿美元规模的母婴市场中,虽然母婴智能设备市场发展快速,智能喂养、发育监测和孕产科技也已成为全球增
资讯
36氪首发 | 大疆前高管押注个人出行机器人,累计融资近亿美元
作者 | 张子怡 编辑 | 袁斯来 硬氪获悉,AI智能硬件与机器人企业「若创科技」(英文名:Strutt)近日完成Pre-A+轮融资,累计融资金额近亿美元。本轮融资获得全球领先出行科技企业、国内头部互联网企业旗下投资基金及大湾区代表性产业基金共同加码。 本轮融资资金将主要用于个人物理AI技术研发、全球营销与销售渠道建设,以及首款个人智能出行机器人EV1的规模化生产和交付。同时,公司正式启用中文品牌“昂首”,首款个人出行机器人在中国市场的官方命名为“昂首EV1”,计划于2026年下半年启动中国市场业务,目前已开放首批试驾报名。 若创科技成立于2023年,总
资讯
Qwen-Image-3.0上线千问AI平台
36氪获悉,据阿里云消息,阿里巴巴千问图像生成模型Qwen-Image-3.0正式上线千问AI平台,面向所有用户全部开放。旗舰版Qwen-Image-3.0-Pro与标准版Qwen-Image-3.0-Standard同时开放API,文生图0.18元/张起。API已在千问AI平台上线,海外用户可通过Qwen Cloud接入。
资讯
京东:开源实时流式视频编辑模型JoyAI-Video-Edit
36氪获悉,据京东黑板报消息,8月5日,京东宣布开源自研的实时流式视频编辑模型JoyAI-Video-Edit。据介绍,用户可以一边观看视频,一边修改人物与场景,让视频创作从“先有素材再修改”变为可实时互动编辑。
资讯
中信证券:传统主业稳健+AI收入渐显,美股应用软件迎来估值修复契机
36氪获悉,中信证券研报表示,当下“AI吞噬软件”叙事正在逐步证伪:龙头厂商收入与在手订单仍保持稳健增长,应用软件基本面尚未出现市场此前担忧的结构性失速;同时模型厂商通过嵌入CRM、ERP、ITSM和HCM等既有工作流推进企业级落地,传统软件厂商的数据、行业知识及客户服务能力仍构成核心壁垒。短期看,模型同质化属性提升、软件企业AI收入占比提升等有望推动美股软件板块估值持续向上修复。中期维度,板块能否由估值修复走向反转,仍取决于AI产品能否带动软件企业整体营收增速回升,未来1—2个季度可能是重要观察窗口。
查看全部 60 条 →
📚 国际技术论文
arXiv · cs.AI / cs.CL / cs.LG
2026-10-18 10:15
论文
LIBERO-PeRM: Benchmarking Personalized Robotic Manipulation
Personalization is essential for general-purpose robots to transition into household environments and achieve true practicality. While general robotic manipulation has been extensively studied, personalized robotic manipulation (PeRM) lacks a controlled benchmark for separating task completion from preference satisfaction. We introduce LIBERO-PeRM, a benchmark for preference learning in embodied a
论文
Discovering state-of-the-art quantum circuits with large language models
Significant efforts are currently being directed toward the implementation of large-scale quantum information processors, but realizing useful quantum advantage also requires the development of efficient quantum algorithms. While for many practical problems the dominant resource costs are associated with a small set of computational primitives, optimizing them in the quantum setting presents uniqu
论文
T2-z: Test-Time Optimization of Environment Latents for Robotic World Models
Predicting how the world responds to an action requires more than what can be seen: robots must account for environmental factors that are neither directly observable nor explicitly specified, from physical properties such as friction and mass to human-defined environmental rules. We introduce T2-Z for test-time optimization of environment latents, which augments a robotic world model with a learn
论文
CoDA: Co-Evolution of Data Engines and Action Policies
Generative simulation can automate robot environment construction, but physically valid environments are not necessarily useful for training. The challenge is to identify environments that address a policy's failures and supervision that helps it overcome them. We introduce CoDA, a framework in which a simulation data engine and an action policy co-evolve through a shared heterogeneous memory of e
论文
PathFM: Fast Generative Forecasting of Multivariate Trajectories
Time series foundation models are becoming increasingly popular for general-purpose forecasting. However, most leading models predict marginal quantiles, leaving the dependence structure of future trajectories unspecified. This distinction matters for trajectory-dependent quantities such as cumulative demand, peak load, and event timing, yet point and marginal scores do not reveal it. To model thi
论文
OmniVerifier-C1: Decomposed Graded Verifier for Rule-Governed Generative World Modeling
Generative visual verifiers play a crucial role in advancing multimodal intelligence. To enhance the robustness of visual verifiers in complex multimodal scenarios involving compositional requirements, we move beyond conventional binary verifiers that produce only True/False discrete judgments and introduce a verifier that automatically decomposes complex prompts or environmental rules into fine-g
查看全部 60 条 →
📖 中文技术文章
InfoQ 中国
2026-10-18 10:15
技术
“AI+终端”时代的萌芽:大模型在终端领域的探索和量产落地|QCon上海
点击查看原文>
技术
Akka 在 65 个开源项目中测试基于规格的 AI 交付
点击查看原文>
技术
The World Is Not a Dataset - Building AI Systems That Continuously Learn from Reality|QCon上海
点击查看原文>
技术
Java 近期新闻:JobRunr 9、OpenXava 8、Quarkus、LangChain4j、JNoSQL、Lathe
点击查看原文>
技术
聚焦 AI 场景落地与 FDE 创新,2026 人工智能产业大会专项活动在济南举行
点击查看原文>
技术
生产级大模型平台工程实战手册
点击查看原文>
查看全部 60 条 →