Models & Labs
Base44 Launches Custom AI Model for Vibe Coding
TechCrunch AIhigh confidence
Why it matters
- →Base44's custom model aims to optimize performance and reduce costs, enhancing its competitive edge.
- →The move reflects a broader trend of AI companies seeking defensibility through proprietary data and infrastructure.
- →Specialization in AI models could offer advantages over general-purpose frontier models.
Base44, a vibe coding platform acquired by Wix for $80 million, is rolling out its own AI model, Base1, to support app creation with natural language. This strategic move aims to optimize latency, cost, and efficiency by integrating the model into its tech stack, differentiating it from competitors like Lovable that rely on external models. The launch reflects a trend among AI companies to leverage proprietary data and infrastructure for defensibility. Base44's approach highlights the growing importance of specialization in a competitive AI landscape.
Read originalMore from TechCrunch AI
© TechCrunch AIMarket & Regulationbusiness
AI Adoption Spurs Job Growth in Tech Firms
A new report challenges the narrative that AI is a job killer, showing that companies heavily investing in AI are actually increasing their workforce. These 'high-intensity adopters' are seeing a 10.2% rise in headcount, particularly in tech-forward sectors like software and media. While AI is often feared for eliminating jobs, this data suggests it can also drive firm expansion, especially in roles like engineering and customer service. However, the benefits are not universal, as firms without sustained AI investments see no headcount gains, highlighting a growing divide in the business landscape.
TechCrunch AI
© TechCrunch AIVideo & Creative AIimage
Gemini's AI Image Generation Now Free in U.S.
Google's Gemini app has expanded its personalized AI image generation feature to all eligible U.S. users for free, previously limited to paid subscribers. This feature, powered by Nano Banana, allows users to generate images based on their personal preferences without specifying them in prompts. By leveraging data from Google services like Gmail and Google Photos, Gemini can create tailored illustrations, even using actual user photos. This move democratizes access to advanced AI tools, making personalized image creation more accessible to a broader audience.
TechCrunch AI
© TechCrunch AIMarket & Regulationbusiness
California Secures Discounted Access to Anthropic's Claude
California has entered into an agreement with Anthropic to provide its government agencies with access to the AI chatbot Claude at a reduced price. This initiative is part of Governor Newsom's strategy to enhance government efficiency by leveraging AI to assist in drafting documents and analyzing data, while ensuring that human roles remain central. The deal aligns with Newsom's executive order to integrate AI into government operations responsibly. This collaboration represents a different approach from the federal government, as Anthropic previously faced challenges with the U.S. Department of Defense over ethical use concerns, resulting in a strained relationship.
TechCrunch AI
More in Models & Labs
Models & Labsmodels
vLLM v0.24.0 Release Enhances Model Support
The vLLM v0.24.0 release marks a significant update with extensive contributions from 256 developers, introducing support for new models like MiniMax-M3 and DiffusionGemma. This version enhances performance with optimizations such as the FlashInfer sparse index cache and improved throughput for DeepSeek-V4. The update also expands the Model Runner V2 capabilities, supporting quantized models by default and integrating GraniteMoE. These advancements make vLLM more robust and versatile, offering developers improved tools for model deployment and performance tuning.
vLLM Releases
Models & Labsmodels
Llama.cpp b9833 Release Enhances MiniCPM5 Parser
The latest b9833 release of llama.cpp focuses on refining the MiniCPM5 parser, addressing several technical aspects to improve its functionality. This update includes the addition of a new tool call parser, refactoring of the PEG parser, and adjustments to the Jinja min/max API for better compatibility with Jinja2. The release also reverts some shared mapper changes to maintain strict JSON parsing for tool-call arguments. These enhancements aim to streamline the parsing process, ensuring more reliable and efficient handling of XML tool calls and grammar triggers.
llama.cpp Releases
Models & Labsmodels
llama.cpp b9835 Release Expands Platform Support
The latest b9835 release of llama.cpp continues its trend of broadening platform compatibility, though without major new features. Notably, the release includes support for ROCm 7.2 on Ubuntu x64, which is significant for AMD GPU users seeking alternatives to NVIDIA's CUDA. The update also maintains a wide array of builds across macOS, Linux, Windows, and openEuler, ensuring developers have the flexibility to deploy on diverse systems. While the release doesn't introduce groundbreaking changes, it solidifies llama.cpp's position as a versatile tool for AI inference across multiple environments.
llama.cpp Releases