Latest AI News for week (07Apr-13Apr2025)

1. Anthropic Introduces Claude Max Subscription Plans

Anthropic has launched new subscription tiers for its Claude AI chatbot, targeting professional users. The “Max” plans offer significantly increased usage limits: the $100/month plan provides 5 times the usage of the Pro plan, while the $200/month plan offers 20 times the usage. Subscribers also gain early access to new features and priority support during peak times.

2. Google Unveils Ironwood TPU v7 for AI Inference

Google has announced its seventh-generation Tensor Processing Unit (TPU), named Ironwood, designed specifically for AI inference tasks. Ironwood delivers peak performance of 4,614 teraflops and supports large-scale deployment in clusters of up to 9,216 chips. This advancement aims to enhance performance and energy efficiency in AI applications.

3. xAI Releases Grok 3 API with Multiple Versions

xAI has officially launched the Grok 3 API, offering both standard and inference-optimized models. Each model is available in regular and fast modes, supporting a context length of up to 128K tokens. The fast mode provides quicker responses at a higher cost, while data-sharing users receive a monthly credit of $150, making it an attractive option for developers.

4. Reallusion Launches Character Creator 5 for Next-Gen 3D Avatars

Reallusion has introduced Character Creator 5 (CC5), enhancing 3D digital human creation with features like 8K ultra-realistic skin textures and AI-driven facial animation. The new system supports detailed micro-expression capture and integrates seamlessly with tools like Maya and Marmoset, marking a significant step forward in digital avatar realism.

5. Kimi Open-Sources Lightweight Vision-Language Model

Kimi has open-sourced its lightweight vision-language model, Kimi-VL, and its inference variant, Kimi-VL-Thinking. Utilizing a Mixture of Experts (MoE) architecture with 16 billion total parameters, only 2.8 billion are active during inference. The model supports a 128K context window and excels in visual understanding, OCR recognition, and complex reasoning tasks.

6. Google Introduces Firebase Studio for AI-Assisted App Development

Google has released a preview of Firebase Studio, a full-stack platform enabling users to create applications through natural language in a browser environment. Integrated with Gemini for AI assistance, it offers 60 templates and supports seamless building, modification, and deployment processes. Early feedback suggests that precise prompts are necessary to achieve optimal results.

7. SenseTime Releases Sixth-Generation Model SenseNova V6

SenseTime has unveiled its sixth-generation large model, SenseNova V6, featuring a 600-billion-parameter MoE architecture. The model integrates text, image, and video processing capabilities, supporting direct reasoning over medium-length videos up to 10 minutes. It aims to bring AI applications closer to everyday scenarios through advanced multimodal understanding.

8. Google Launches Agent2Agent Protocol for Cross-Platform AI Collaboration

Google has introduced the open-source Agent2Agent (A2A) protocol, facilitating secure communication and collaboration between AI agents across different platforms and vendors. A2A encompasses core functions like capability discovery, task management, collaboration, and user experience negotiation, supporting long-running tasks and multimodal interactions.

9. White Paper Advocates AI-Driven Open Science in Astronomy

The Chinese Academy of Sciences and Zhejiang Lab have published a white paper promoting the use of AI to empower open science in astronomy. It addresses challenges in storing, analyzing, and sharing petabyte-scale astronomical data and proposes collaborative efforts like the GOTTA telescope array and the establishment of the “GOTTA Academy” for intelligent education.