The Week AI Went All Out 50% off Google AI Pro, 2x Claude limits, GPT-5.2-Codex, Gemini 3 Flash, and Nvidia's $20B power move
Sponsored by
Create full-stack, production-ready web and mobile apps from natural language prompts with Emergent — the first agentic vibe-coding platform that handles everything from planning to deployment.
Learn MoreHappy Boxing Day! 🎁 While you're recovering from the holiday rush, AI companies decided to go all out: Google slashed AI Pro by 50%, Anthropic doubled Claude limits through New Year's, OpenAI dropped GPT-5.2-Codex, Gemini 3 Flash went live globally, Nvidia acquired Groq for $20 billion, and MCP got donated to the Linux Foundation. This wasn't a slow news week. This was the week that redefined the landscape.
On this day in 2024, Google unveiled Willow, a quantum chip that cracked a 30-year error correction challenge and performed calculations in 5 minutes that would take traditional supercomputers 10 septillion years. One year later, we're watching AI models evolve at quantum speed.
Let's dive into what happened while Santa was making his rounds. 🎅
🎉 Google slashes AI Pro plan by 50% in New Year's promotion
Google is running a limited time deal for new subscribers, cutting annual Google One plans in half through New Year's Eve. The AI Pro tier drops from $199.99 to just $99.99 for the first year, bundling 2TB storage, higher Gemini 3 Pro access, Deep Research, Workspace AI integration, and perks like 10% Google Store credit. The 2TB storage plan is also discounted to $49.99 (normally $99.99). The catch: new subscribers only, and prices return to normal after year one. Sign up at one.google.com/ai-nye.
🎁 Anthropic doubles Claude usage limits for Pro and Max subscribers through New Year's Eve
Anthropic is giving Pro and Max plan subscribers a holiday gift: 2x usage limits from December 25 through December 31. Both five-hour session limits and weekly caps are doubled for the promotion. The boost applies to claude.ai, Claude Code, and Claude in Chrome. Individual subscribers only (Team and Enterprise plans aren't included). New subscribers who sign up during the promotional period are also eligible. No extra charge, it's just a thank you to paid users for the holidays.
🚀 OpenAI releases GPT-5.2-Codex, their most advanced agentic coding model
OpenAI launched GPT-5.2-Codex, a version of GPT-5.2 optimized specifically for their Codex coding agent. The model brings better long-horizon work through context compaction, stronger performance on large code changes like refactors and migrations, improved Windows support, and significantly enhanced cybersecurity capabilities. A security researcher using the previous version already discovered and disclosed a vulnerability in React. GPT-5.2-Codex is available now for paid ChatGPT users across all Codex surfaces, with API access coming in the next few weeks after additional safety testing.
⚡ Google launches Gemini 3 Flash with Pro-level reasoning at fraction of the cost
Google released Gemini 3 Flash, combining Gemini 3 Pro's reasoning capabilities with Flash-level speed and efficiency. The model scores 90.4% on GPQA Diamond and 78% on SWE-bench Verified (outperforming even Gemini 3 Pro on coding), while running 3x faster than 2.5 Pro at just $0.50/1M input tokens. It's now the default model in the Gemini app for all users globally and rolling out in AI Mode in Search. Developers can access it via Google AI Studio, Gemini CLI, Vertex AI, and Google Antigravity.
💰 Nvidia acquires Groq for $20 billion in largest-ever deal
In a stunning move announced December 24, Nvidia is acquiring AI chip startup Groq's assets for approximately $20 billion — its largest purchase on record. The nine-year-old chip startup has been building specialized inference accelerators, and this acquisition signals Nvidia's aggressive push to dominate both training and inference markets.
🤝 Anthropic donates Model Context Protocol to Linux Foundation
Anthropic donated the Model Context Protocol (MCP) to the Linux Foundation through the newly established Agentic AI Foundation, co-founded by Anthropic, Block, and OpenAI, with backing from Google, Microsoft, AWS, Cloudflare, and Bloomberg. The move ensures MCP remains vendor-neutral and community-driven as it becomes critical infrastructure for AI agents. The protocol has already seen 97M+ monthly SDK downloads across Python and TypeScript.
💬 Chatterbox is Resemble AI's state-of-the-art open-source TTS model that's taking GitHub by storm (15.9K stars). The latest Chatterbox-Turbo model packs a streamlined 350M parameters and supports 23 languages with emotion exaggeration control and paralinguistic tags like [laugh], [cough], and [chuckle]. Every audio file includes imperceptible neural watermarks that survive MP3 compression. It's the first open-source TTS to offer this level of control — and it's MIT licensed.
🖥️ exo lets you run your own AI cluster at home using everyday devices — phones, laptops, desktops, even smartwatches. Devices running exo automatically discover each other with zero manual configuration. With day-0 support for RDMA over Thunderbolt 5, it achieves 99% reduction in latency between devices. Updated as recently as December 23, 2025, this project is democratizing AI infrastructure that typically requires expensive GPU servers.
⚡ LEANN is the groundbreaking open-source RAG framework that's exploding on GitHub (over 6K stars and rapidly climbing). Developed by yichuan-w, LEANN delivers 97% storage savings through innovative compression techniques while enabling fast, accurate, and fully private Retrieval-Augmented Generation on personal devices—no cloud required. It supports running sophisticated RAG pipelines locally with minimal resources, making advanced AI accessible on laptops or edge hardware. Fully open-source and privacy-focused, LEANN is redefining efficient, offline AI applications for developers and researchers alike.
Coming in Hot
Affordable AI gateway with unified access to leading AI models at lower costs.
AI trading journal that builds discipline habits.
AI video generator with multi-shot storytelling
AI humanizer that transforms robotic text into natural writing
AI-powered music video editor with beat synchronization.
AI video repurposing for social media clips.
Turn voice notes into polished content with AI.
AI sales agent that sells for you on WhatsApp & Instagram.
AI meeting recorder that turns calls into insights.
Advanced AI image generation and editing models.
Submit your AI tool to the #1 AI tools marketplace.
What it does: Appy Pie offers a no-code AI app generator with a freemium model; pricing plans and integration costs vary. Reviews indicate issues with service delivery and billing. Current pricing details are available on the Appy Pie Automate pricing page.
Key features:
Pricing: Freemium model with multiple tiers. The platform offers free options with limited features, while paid plans unlock professional app development services and additional integrations. Appy Pie Automate has tiered pricing for workflow automation.
Recent updates: The AI App Generator continues to receive updates with enhanced AI guidance features. However, recent Trustpilot reviews from January 2025 highlight concerns about service delivery timelines and billing practices that potential users should consider.
Why it matters: Appy Pie democratizes app development for non-technical users, enabling entrepreneurs and small businesses to build digital products without hiring developers. The AI-powered approach significantly reduces the time and cost barrier to entry for mobile app creation.
What it does: Mindgrasp helps you learn faster by turning any content into summaries, notes, quizzes, and more using AI.
Key features:
Pricing: Offers both individual and business/enterprise solutions. Simple and transparent pricing structure designed for students and corporate training needs. The platform transforms corporate documents, meetings, and training materials into strategic assets and actionable insights for business users.
Recent updates: Mindgrasp has expanded its features to target specific student demographics like accounting students with specialized study tools. The platform continues to refine its flashcard generation and summary capabilities based on user feedback showing improved academic performance.
Why it matters: Mindgrasp addresses the growing need for efficient learning in an information-overload era. Students and professionals dealing with massive amounts of content can compress hours of reading into minutes of focused study. The AI-powered flashcards and quizzes provide active learning mechanisms that improve retention far beyond passive reading.
What it does: ModelsLab offers competitive pricing for AI models, with Flux models ranging from $0.004 to $0.08 per generation. Vertex AI pricing varies from $0.02 to $0.06 per image. Gemma 3n E4B and DeepSeek-OCR are among the cheapest models at $0.03 and $0.05 per million tokens.
Key features:
Pricing: Transparent pay-per-use model: FLUX.1 Pro at $0.07/generation, Flux Pro 1.1 Text To Image at $0.06/generation, Flux Pro 1.1 Ultra at $0.08/generation. Imagen 4 Ultra image generation at $0.06/image with upscaling at the same rate. This aggregated pricing model offers significant cost savings compared to integrating each provider's API individually.
Why it matters: ModelsLab solves a critical pain point for developers: managing multiple AI provider accounts and APIs. By aggregating access to leading models from BFL, Google, and others under one roof, it simplifies infrastructure while reducing costs. For startups and indie developers building AI-powered products, this unified API approach means faster development cycles and more predictable expenses.
🛡️ New York passes RAISE Act for AI safety oversight
Governor Hochul signed the RAISE Act on December 19, requiring large AI developers to publish safety protocols and report incidents within 72 hours. The law creates an oversight office within the Department of Financial Services and builds on California's framework to create a unified benchmark among leading tech states.
💸 Resolve AI hits $1B valuation building autonomous SRE
Ex-Splunk executives' startup Resolve AI raised a Series A led by Lightspeed Venture Partners at a $1 billion valuation. The company is developing an autonomous site reliability engineer (SRE) that automatically maintains software systems — addressing the growing complexity of modern infrastructure.
🎯 Zoom unveils AI Companion 3.0 with agentic workflows
Zoom launched AI Companion 3.0 on December 15 with agentic features that automatically execute follow-up tasks. Personal workflows can compile insights from meetings and documents to deliver daily reflection reports or automatically summarize Team Chat threads — transforming conversations into action without manual intervention.
🧠 Yann LeCun's new AI startup targets $3.5B valuation before launch
Less than a month after leaving Meta, Turing Award winner Yann LeCun launched fundraising talks valuing his new venture, AMI Labs, at roughly $3.5 billion — before it's even launched. The 65-year-old AI pioneer's move signals major developments ahead in fundamental AI research.
📊 MIT Technology Review declares 'great AI hype correction of 2025'
A new MIT Technology Review analysis examines how 2025 marked a pivotal shift from AI hype to practical deployment challenges. As models reach new capabilities, the industry is grappling with the gap between demo performance and production reliability — forcing a more realistic assessment of AI's current limitations.
🎓 Cornell study: AI boosts research output but floods journals with mediocre papers
Cornell researchers found that while AI tools help scientists work faster, they're also contributing to an explosion of lower-quality submissions. Major AI conferences are now turning down already-approved papers to manage volume, revealing tensions between AI-assisted productivity and research quality standards.
That's it for this week. Hope you found something useful in here. If you did, send it to someone who'd appreciate it. 🎉
The AI race is hitting peak velocity, and we're here to make sure you don't miss a thing. See you next week with more breakthroughs, funding rounds, and tools that actually matter.
Follow us on Threads, X, and Instagram for more AI updates.
— Somi AI
