This matters because it offers a cheaper alternative for building AI agents, with a quarter of the memory needs of its predecessor.
Deepseek and Alibaba release trillion-parameter models, OpenAI launches Agents API and financial ChatGPT
2 Sep 2026. Trillion-parameter models dominate today's news as Deepseek releases V4.1-Flash with 552 billion parameters and Alibaba publishes Qwen3.8-2.4T-A95B, both pushing the boundaries of large language model scale. OpenAI meanwhile expands its capabilities with the public launch of Agents API for multi-step task execution and a new financial-focused ChatGPT service powered by GPT-6 Astra. Industry developments include Ant Group's role in AI agent ecosystems and Kimi's enterprise partner program, while research highlights new memory management systems and unlearning frameworks for LLMs.
Models
3 itemsFirst open-source MoE model matching Qwen-Max performance, offering developers enterprise-grade capabilities without cost.
This matters because it's a new, more efficient 552B model from DeepSeek that reduces hardware requirements and costs, making it more accessible for developers and businesses.
Products
3 items亚马逊这个新功能能帮你更快地得到 LLM 的回答,特别是处理大量相似问题的时候,比之前的方法效率高很多。
OpenAI 推出了 Agents API 公测版,开发者可以通过 API 调用云端 AI 智能体运行环境,支持多步骤任务运行,还能拆分任务让多个智能体并行处理,挺实用的。
OpenAI 推出了面向金融服务的 ChatGPT 版本,集成 GPT-6 Astra 和金融数据,能帮你处理财报、做估值模型,比普通版本更专业。
Industry
3 itemsAnt Group's CEO reveals that Alipay will launch incentive policies to support developers in creating useful agents and connect user needs with merchant services.
For users interested in enterprise AI solutions, this program offers a way to integrate Kimi's capabilities with established tech partners.
For readers in China, this story clarifies the status of Huaqiangji's pricing, which is relevant to those considering their AI chip purchases.
Papers
3 itemsFR-Bank improves memory accuracy in LLMs by 12.7 percentage points over Mem0, directly benefiting developers building reliable AI applications.
FOM-UL improves privacy compliance for LLMs while preserving utility, crucial for companies deploying models with sensitive data.
How-to
3 itemsFor developers using LangChain agents, this shows how to scale evaluation beyond manual checks.