News linked to both this project and an event.
According to The Wall Street Journal, following Moonshot AI's release of the low-cost, high-performance large model Kimi K3, Wall Street has once again seen concerns about a "DeepSeek-style shock," with the AI chip sector facing selling pressure. Morgan Stanley stated that the market may experience AI industry turbulence similar to that triggered by DeepSeek in early 2025, where the rise of low-cost, high-performance AI models could challenge top U.S. model developers such as OpenAI and Anthropic, while simultaneously weakening demand for the large-scale compute infrastructure underpinning the U.S. AI investment boom.
According to Bloomberg, AI company Moonshot AI has reportedly informed investors that the company plans to go public within six months at the earliest. Sources revealed that Moonshot AI has officially distributed shareholder resolution documents to investors, seeking shareholder support for its listing in Hong Kong. Initiating this notification process means that the company could complete the IPO within the next six months at the earliest. Moonshot AI is currently completing a round of financing, and the company's valuation could exceed $30 billion post-financing. They stated that the company believes now is the right time to go public because its Annual Recurring Revenue (ARR, a key metric measuring future sales capability) reached $300 million in June.
analyst KawzInvests stated that Moonshot AI's upcoming Kimi K3 could become a significant event in the open-source AI space, and the infrastructure demand behind it may drive growth for AI cloud service platforms. Kimi K3 has approximately 2.8 trillion parameters, making it one of the ultra-large-scale open-source models. According to Moonshot's official evaluation, the model's performance is only slightly behind frontier models like Claude Fable 5 and GPT 5.6 Sol, and the full model weights are expected to be released on July 27.KawzInvests pointed out that a model of this scale cannot run on an ordinary laptop or even a single server; users need a computing cluster composed of a large number of GPUs to complete model loading and inference. When top-tier open-source models are made available for free, the biggest beneficiaries might not be ordinary users, but rather platforms that offer model hosting and inference services. For example, $DOCN (DigitalOcean) already supports serverless inference services for models like Kimi K2.6. Developers do not need to deploy hardware; they can call the model via API and pay per Token. Additionally, the platform hosts over 70 models and covers GPU leasing, model fine-tuning, and AI Agent development tools.As more large-scale open-source models are released, developers' demand for low-barrier AI infrastructure will continue to increase. Model hosting, inference services, and GPU cloud platforms may become key beneficiaries in the open-source AI wave.
: Yesterday, Dark Side of the Moon (Moonshot AI) released its latest open-source AI model, Kimi K3. It ranked first on the Frontend Code Arena test website with a score of 1,679, surpassing the Claude Fable 5 model. Following an evaluation of the K3 model by Artifacial Analysis, Elon Musk once again praised the Kimi model from Dark Side of the Moon, stating that the K3 model's benchmark performance is impressive.In March of this year, when Kimi published the research paper "Attention Residuals: Rethinking the Aggregation of Depth Direction," it received praise from Musk, who said, "Kimi's research work is impressive." Previously, he also stated that the Zhipu GLM model could surpass the Claude Mythos model (i.e., Fable 5) by Q1 2027. In response, Zhipu founder Tang Jie replied, "It won't take that long."
on July 17 that B.AI announced the official launch of the Kimi K3 model on its API platform, achieving seamless integration shortly after the release by @Kimi_Moonshot. As the world's first open-source 3T-level (2.8 trillion parameter) model, Kimi K3 has made significant breakthroughs in long-range programming, complex reasoning, and knowledge work, while natively supporting visual capabilities and 1M ultra-long context. As a model aggregation service platform, B.AI has prioritized the opening of API interfaces this time, allowing developers to directly experience the new generation of model capabilities through B.AI. B.AI will continue to promote the rapid implementation of cutting-edge AI technology and empower developer innovation through efficient and stable interface services.
According to Xinhua News Agency, Beijing Moonshot AI Co., Ltd. officially released the new generation large model Kimi K3 on July 16, with a parameter scale reaching 2.8 trillion, becoming the largest open-source model by parameters globally at present. Kimi K3 natively supports visual understanding, features a 1 million token context window, and is optimized for complex task scenarios such as software engineering, knowledge work, deep research, and multimodal understanding, with a comprehensive intelligence level approaching global frontier closed-source models. The model is trained using a self-developed underlying architecture. Moonshot AI stated that nearly 3 trillion parameters mean the model can store more knowledge and patterns, achieving "knows more, thinks deeper, answers more accurately."
: Fintech platform MoonPay has acquired crypto infrastructure startup Glide and will integrate its deposit and routing technology. Financial terms of the transaction were not disclosed. Glide was founded in 2023 by Tushar Soni and Qinyu Tong, both of whom were part of the Robinhood Wallet team. Glide supports over 100 tokens across 30 blockchain networks, helping applications receive deposits from various tokens, wallets, exchanges, and payment sources. Upon completion of the acquisition, Glide's technology will be integrated into MoonPay Deposits, a product already used by applications such as Wallet in Telegram, Moonshot, and Paysafe. MoonPay stated that this transaction is part of its efforts to expand its digital asset infrastructure capabilities. This is the sixth acquisition announced by MoonPay in 2026, following previous acquisitions of Sodot, Decent, DFlow, Entendre, and Dawn Labs. Its investors include Thrive Capital, Paradigm, Valhalla Ventures, Tiger Global Management, and Coatue.
The PPP Prediction Market tool monitors that Polymarket has listed a new event: "Will the US government revoke public access to another major AI model?" The current probability is reported at 33%.The settlement rules state: If the US federal government, by the end of 2026, passes relevant legislation, issues an executive order, implements export controls, or takes any other action that substantially restricts US public access to a major AI model, the market outcome will be "Yes." Otherwise, the outcome will be "No." A "qualifying action" refers to a formal measure taken by the US government, the effect of which is equivalent to completely prohibiting the public from accessing a specific AI model within the United States. Furthermore, the settlement rules emphasize that regardless of the action's true purpose or nominal goal; if the action effectively results in the public being unable to access the model within the US—for example, prohibiting the model from being provided to foreign citizens or governments, as long as the general public cannot access the model through conventional channels within the US—it meets the qualification requirements. Merely excluding access to the model from a single channel is insufficient. Removals of public access that are not caused by any formal action of the US government are not eligible."Mainstream AI model" refers to the flagship, general-purpose large language model or multimodal foundation model developed by one of the following companies: OpenAI, Anthropic, Google (including Google DeepMind), Meta, xAI, Microsoft, Amazon, Mistral AI, DeepSeek, Alibaba, ByteDance, Moonshot AI, and Zhipu AI (Z.ai). Models designed for specific tasks, or those that are outdated, used solely for research or preview purposes, do not meet this criterion.The action can target a single model or a group of models, as long as at least one major AI model becomes inaccessible to the public within the US as a result. A temporary suspension of public access to a model meets this condition. However, if an action has been implemented or a related resolution has been issued, but the public can still access the model before the resolution takes effect, that action does not meet the condition.The information sources for this market are official announcements and information from the US government and the relevant AI companies. However, reliable media reports may also be referenced to form a consensus.The Odaily Seer Channel continues to monitor prediction markets, observing changes before prices are set.
: DeepSeek has officially launched the image recognition mode on its web version, while the app is still in the internal testing phase. Functionally speaking, the newly launched image recognition mode is still in its early stages. The reporter repeatedly tested the model using photos of DeepSeek founder Liang Wenfeng. After several minutes of deliberation, the model gave many incorrect answers and ultimately admitted, "I am really not sure. It is not recommended to guess the name because guessing wrong is worse than saying 'I don’t know.'" In other tests by netizens, when presented with a photo of founder Liang Wenfeng, DeepSeek even misidentified him as Yang Zhilin, founder of Moonshot AI, or a "younger version" of Ma Huateng.It is worth noting that just before the launch of the image recognition mode, DeepSeek was reported to have secured its first round of financing, exceeding 50 billion yuan. According to reports, the core investment terms for DeepSeek are: whether for major tech companies or VC funds, Liang Wenfeng’s most important requirement is: not to poach DeepSeek’s employees or suggest they leave to start their own ventures. The reporter inquired about the relevant situation with an insider from DeepSeek and learned that this account is largely accurate. (The Paper)
According to PR Newswire, MoonPay has officially launched MoonPay Headless Onramps—the industry’s first native crypto checkout platform enabling one-click crypto purchases via Apple Pay, credit cards, and Google Pay across the U.S., the European Economic Area, and over 100 countries worldwide. In contrast, competing headless payment solutions currently support mobile payments only within the U.S. This product replaces MoonPay’s branded widget with a pure API integration, enabling partners to deliver a fully white-labeled and highly customizable checkout experience—while MoonPay handles payment processing, compliance, and identity verification in the background. Launch partners include Moonshot, Bitcoin.com, Bread, and Trust Wallet. Notably, Apple Pay is now fully embedded into partner apps for the first time: verified users can complete purchases with a single tap—no redirects or re-verification required.
According to LatePost, Kimi, a product of Moonshot AI, is about to complete a new round of financing worth $2 billion, pushing its post-money valuation beyond $20 billion. This round is led by Meituan Longzhu, with participation from China Mobile and CPE Capital; Meituan Longzhu’s investment exceeds $200 million. The report notes that, including three financing rounds already completed in January and February this year, Kimi has raised over $3.9 billion in less than six months—more than RMB 3.76 billion in total—making it one of the most well-funded large-model startups. The article also mentions that in April, Kimi released and open-sourced its new model K2.6, enhancing its programming and Agent cluster capabilities.
Moonshot AI has released its new model, Kimi K2.6, which is now live on kimi.com in both chat and Agent modes. According to the official announcement, the model achieves top-tier performance across multiple open-source coding and tool-use benchmarks, including HLE with tools, SWE-Bench Pro, SWE-bench Multilingual, and BrowseComp. Meanwhile, Kimi K2.6 supports over 4,000 tool calls, continuous execution for more than 12 hours, and handles multi-language tasks—including Rust, Go, and Python. Its parallel Agent capability has been enhanced to support up to 300 concurrent sub-Agents and 4,000 steps per run, and is already deployed in autonomous operation scenarios such as OpenClaw and Hermes Agent.