A (Last) Week in AI: Major Trends and Highlights

By on 21 April 2025
in News

Exciting Leaps in Video, Coding, and Next-Gen Models
By Mulubwa Chungu – Technical Lead, BongoHive Consult | Gen AI Core Team Support Lead

This past week marked a major turning point in artificial intelligence. From video generation breakthroughs to next-gen models and enhanced productivity tools, tech giants like Google, OpenAI, Anthropic, Microsoft, and ByteDance are pushing the limits of what AI can do.

🔥 Google’s Gemini 2.5 Flash Sets a New Standard

Google launched Gemini 2.5 Flash, a lightweight, high-speed model designed for efficiency and scale. It’s now ranked joint #2 on LMSYS’s Chatbot Arena, alongside GPT-4.5 Preview and Grok-3. Best of all? It’s free to try.

🤖 OpenAI Rolls Out ChatGPT 4.1, o3, o4-mini Models, and a New Coding Agent

OpenAI dropped multiple updates:

  • ChatGPT 4.1, o3, and o4-mini models
  • A new autonomous coding agent that builds and deploys full apps from natural language prompts — a huge step toward low-code/no-code AI development.

🧠 Grok AI Gets Studio & Memory Features

Elon Musk’s AI assistant Grok now includes:

  • Grok Studio: Create your own custom chatbots
  • Memory: Stores previous conversations, similar to ChatGPT
    More on that here.

🧪 Claude by Anthropic Moves Toward Autonomous Research

Anthropic is equipping Claude for independent research and insight generation, showing serious promise in autonomous task handling and long-term memory applications.

🎥 Kling 2.0 Redefines AI Video

Kuaishou’s Kling 2.0 stunned the AI world with ultra-realistic videos, showcasing complex human movement and cinematic camera control. SCMP called it “the most powerful AI video generator” yet, rivaling OpenAI’s Sora.

🎨 Canva’s Visual Suite 2.0 Supercharges Creativity

Canva unveiled its biggest upgrade yet:

👀 Microsoft Copilot Upgrades Vision

Microsoft’s AI assistant now interprets images, charts, and full multi-modal documents — a leap in workplace productivity and data comprehension.

📹 ByteDance Releases Seaweed AI

TikTok’s parent company introduced a lightweight video model, Seaweed AI, built for smoother motion and realistic environments. China’s video AI race is officially heating up.

🌐 FireCrawl’s FIRE-1 Agent Supercharges Web Research

FIRE-1 autonomously browses the web and summarizes content — imagine having a mini researcher in your browser.

🖼️ Freepik Adds Kling AI & Layout Intelligence

Freepik now supports Kling-generated video content and introduces Composition Reference, a layout assistant for storytelling and better design flow.

🐬 Google’s DolphinGemma Makes AI Run On-Device

Optimized for edge devices, DolphinGemma is a compact, high-performance LLM that can run without cloud infrastructure.

📊 Google Sheets Gets Smarter

New AI features in Google Workspace automate data classification, project tracking, and generate actionable insights directly in Sheets.

🎬 Alibaba’s First-Last-Frame Model Changes the Video Game

Alibaba’s new model generates entire video sequences from just the first and last frames — making storytelling faster and more efficient.

What This Means

AI innovation is exploding across industries — from Google’s Gemini Flash and Claude’s research breakthroughs to Canva’s creative reinvention and OpenAI’s coding revolution. Kling 2.0 and Seaweed AI prove video generation is advancing fast, and tools like Copilot, FireCrawl, and Sheets AI are embedding intelligence into our daily workflows.

The takeaway? AI is no longer on the sidelines — it’s woven into the tools we use, the content we create, and the way we solve problems. And it’s only just beginning.

📩 Want to explore AI for your organisation? Let’s talk. Reach out at [email protected]