News & Updates
June 16, 2026
OpenRouter launches Fusion API: "Putting a good mold" to achieve Fable level 5 intelligence at half the cost
OpenRouter launches the Fusion API multi-model collaboration system. The Gemini 3 Flash+Kimi K2.6+DeepSeek V4 Pro combination achieves close performance at about half the cost of Claude Fable 5. The number of OpenRouter access models exceeds 400.
Source: IT Home
Qichacha confirmed: DeepSeek completed the first round of external financing of 51 billion yuan with a valuation of 400 billion yuan
Qichacha data shows that DeepSeek completed its first round of external financing of approximately 51 billion yuan on June 16, with a valuation of nearly 400 billion yuan. Investors include Liang Wenfeng, Tencent, CATL, NetEase, JD.com and the National AI Fund.
Source: IT Home
The Information confirms: DeepSeek completed over US$7 billion in financing, investors have no voting rights and no board seats
The Information reported that DeepSeek completed more than US$7 billion in financing (the largest single round in China's AI industry), adopting an unusual structure: investors have economic rights but no voting rights, no board seats, and a five-year lock-in period. The National AI Fund is the only exception.
Source: IT Home / The Information
June 15, 2026
Google's own Android benchmark: Gemini costs 17.5 times as much as DeepSeek V4 Flash
Google Android Bench list shows that the single running cost of Gemini 3.5 Flash is $147, while DeepSeek V4 Flash is only $8.4, a price difference of 17.5 times. GPT-5.5 ranked first with a score of 74, and domestic models GLM, Kimi, and DeepSeek collectively made the list.
Source: IT Home
June 18, 2026
Doubao vs DeepSeek 2026 Selection Guide: Pan-entertainment full-end collaborative competition for high-end R&D privatization
36Kr AI Evaluation released the 2026 Double Power Selection Guide, which compared Doubao and DeepSeek in depth from the six dimensions of multi-modal creation, logical reasoning, contextual windows, ecological integration, cost, and compliance. Doubao leads the full-scenario multi-modal and traffic ecology, and DeepSeek ranks first in complex inference and private deployment.
Source: 36 krypton (AI evaluation)
It is reported that DeepSeek asked investors to "not poach people", which is the pain behind the loss of core talents.
According to CNBC, DeepSeek founder Liang Wenfeng made special requests to potential investors during the financing process to "not poach DeepSeek employees." Previously, V3 core developer Luo Fuli had switched to Xiaomi.
Source: IT Home/CNBC
June 5, 2026
Have US companies been suffering from high AI prices for a long time? DeepSeek ranks No. 1 on U.S. enterprise software trends list
DeepSeek tops Ramp's software trends list for the first time, as U.S. companies seek alternatives to Chinese models due to the high cost of AI. A company paid Claude $500 million a month, and Uber used up its entire Token budget in four months.
Source: 36 krypton
Doubao must be charged: the end of the free AI era, a turning point in the commercialization of C-side large models
Doubao sparked widespread controversy after announcing the launch of paid subscriptions. In-depth analysis shows that the average daily Token calls exceed 120 trillion and the annual computing power expenditure exceeds 30 billion yuan in structural losses, making the free model unsustainable.
Source: Slow play (authorized release by 36Kr)
DeepSeek's valuation soared to nearly US$60 billion, and China's Internet giants split into the "investment group" and the "self-research group"
DeepSeek’s valuation soared from 45 billion to nearly 60 billion US dollars in one month. Tencent/JD/NetEase formed the "investment faction" and actively entered the game, while Alibaba/Byte/Baidu adhered to the "self-research faction" and watched with cold eyes. Chinese Internet companies split into two camps due to DeepSeek financing.
Source: 36 Krypton (ZAKER Finance)
June 7, 2026
DeepSeek V4 makes mathematical proof: Princeton team completed $170,000 task for $294, a 500-fold cost advantage
The Princeton team used DeepSeek V4 Flash to build Goedel-Architect and completed PutnamBench for $294 (previously requiring $170,000), with a pass rate of 75.6%, surpassing Hilbert. MiniF2F cleared all 244 questions for the first time.
Source: 36 krypton (reprinted from the heart of the machine)
June 11, 2026
AI college entrance examination essay writing competition finals: DeepSeek V4 narrative essay won the championship with 46 points, Hunyuan's "own people" scored full marks
GPT-5.5/Fable-5/DeepSeek V4/Hunyuan 3 Preview Co-written in the 2026 Beijing College Entrance Examination essay, DeepSeek V4's narrative essay won the championship with an average score of 46 points. Hunyuan's "one's own people judge one's own people" scored a perfect score and attracted heated discussion.
Source: 36 krypton (letter AI)
Anthropic launches Claude Corps National AI Philanthropy Fellowship, investing $150M to train 1,000 people
Anthropic spent $150M to launch Claude Corps: 1,000 AI philanthropy fellows with an annual salary of $85K and are assigned to 400+ US non-profit organizations for 12 months.
Source: Anthropic official announcement
Anthropic launches Claude Corps: investing US$150 million to train 1,000 AI Philanthropy Fellows
Anthropic announced the Claude Corps National AI Philanthropy Fellowship, investing US$150 million to train 1,000 Fellows to be stationed in 400+ US non-profit organizations with an annual salary of US$85,000 for 12 months.
Source: Anthropic official announcement
June 4, 2026
AI Vulnerability Challenge actual test: GPT-5.5 has the highest success rate, and DeepSeek V4 Pro costs only 1/15 of it
Security researchers tested multiple AI models with APKs that deliberately left loopholes: GPT-5.5 had the highest success rate of 70%. DeepSeek V4 Pro’s single success cost was only $0.62, which was 1/15 of GPT-5.5. Claude interrupted the session due to security guardrails.
Source: IT Home
DeepSeek's commercial "coming of age" ceremony: from refusing financing to learning to charge
DeepSeek is undergoing a transformation from "not short of money" to "learning to make money" - cutting prices four times a month to grab B-side users, while facing pressure from C-side charges. Doubao has launched paid subscriptions, and DeepSeek's "coming-of-age gift" is to put commercialization on the agenda.
Source: 36 krypton (alphabet list)
June 9, 2026
Cost moat or commons? DeepSeeks cut costs by 99%, where will the US giant’s trillion-dollar bet go?
When open-weighted models such as DeepSeek cut AI costs by 99%, U.S. capital’s trillion-dollar bet based on monopoly assumptions faces a fundamental challenge—AI capabilities are being commoditized.
Source: 36 Krypton (Divine Translation Bureau)
May 28, 2026
A major national fund crosses borders for the first time in 12 years since its establishment: it leads the investment in DeepSeek, and AI enters the "national team-led" stage
The National Fund led the first round of financing in DeepSeek, marking the first cross-border investment in a pure large model company in 12 years since its establishment. The vice president of the China Society of Business Economics commented that the AI industry is entering a stage of "leading the national team and following up with market capital".
Source: IT Home / Financial Magazine
June 17, 2026
Anthropic Seoul office opens: Claude Code, all members of NAVER, Samsung/LG join the Korean AI ecosystem
Anthropic’s Seoul office opened, thousands of NAVER engineers deployed Claude Code, Samsung SDS/LG CNS/Hanwha joined the ecosystem, and signed an AI security MOU with the South Korean Ministry of Science and Technology.
Source: Anthropic official announcement