AI Daily
TOOLIN DAILYMoonshot AI and Tencent Open Source Multimodal Large Models, OpenAI Acquires Development Tools Team
Open-source multimodal and vision models were released simultaneously today, data center construction faced setbacks in multiple regions, enterprise AI deployment spending increased significantly, and multiple organizations launched new robotics and biological research results.
The essentials
Moonshot AI Open Sources Kimi K3.1 Pro Model
Moonshot AI officially open-sourced the Kimi K3.1 Pro model weights and listed them on Hugging Face, featuring a total of 2.8T parameters and 104B active parameters. The model originated from Kimi K3 and underwent reinforcement learning across hundreds of downstream domains, positioned as a native multimodal agentic model, with Ollama Cloud providing a U.S. hosted and zero data retention version.
Source ↗Tencent Hunyuan Releases Hy4 Official Version and Introduces Vision
Tencent released the official Hunyuan Hy4 version, expanding pre-training and post-training scales compared to the preview version, and introduced vision capabilities and computer operation support for the first time. The company stated that GameDev and GameLogic evaluations improved by 19.1% and 9.8% respectively, and it is now open for trials.
Source ↗OpenAI Acquires Development Tools Company Pierre
OpenAI announced the acquisition of Pierre Computer Company, whose Trees and Diffs libraries have been used for file browsing and code change reviews in the Codex application. After the transaction, Pierre's open-source projects will remain open source, and features will be gradually integrated into Codex, pending regulatory approval.
Source ↗In brief
Enterprise AI Deployment Spending Growth
A Menlo Ventures report stated that enterprise AI deployment spending increased from $33 billion in the previous year to $177 billion.
Claude Analyzes Cell Structure
Anthropic's biology team used Claude to analyze the three-dimensional structure of the first prokaryotic vault particle.
ChatGPT Search Volume Share
OpenAI announced that search queries run through ChatGPT have reached 20% of Google search volume.
Claude Code Usage Optimization
Claude Code background sub-agent state updates are now generated by Haiku, significantly reducing usage costs.
Multiple States Reject Data Center Proposals
At least 38 out of over 40 data center-related proposals across 13 states were rejected in the U.S. midterm elections.
OpenAI Releases Images 2.6
OpenAI announced that Images 2.6 is available in ChatGPT, while Codex is open to all tiers simultaneously.