DeepSeek V3, Google Releases Gemini 2.5 Pro, OpenAI Goes Full Ghibli, Microsoft Announces Researcher & Analyst
AI Caps March 2025 with a Bang: DeepSeek's V3 Upgrade, Gemini 2.5 Pro Shines, Microsoft Unveils Researcher and Analyst, & ChatGPT 4o’s Ghibli Twist
DeepSeek Releases AI Upgrade with V3
This week, Chinese AI firm DeepSeek released their V3 model, which has rocketed to the top of performance charts, surpassing competitors in reasoning and multimodal tasks.
V3’s capabilities include autonomous decision-making based on real-time data, like sorting resumes or analyzing stock trends.
The pricing is an incredible deal relative to their peers
Google Releases Gemini 2.5 Pro
Google made waves on March 25 with the launch of Gemini 2.5 Pro, a next-generation reasoning model designed to tackle complex problem-solving with enhanced "thinking" capabilities.
TechCrunch reported the release, noting its potential to rival top models in natural language understanding and logical inference.
3. OpenAI’s ChatGPT 4o Update Goes Full Ghibli
OpenAI rolled out a whimsical update to ChatGPT 4o this week, integrating a Studio Ghibli-inspired aesthetic into its multimodal outputs. Grant Slatton may have started the trend with this post.
The update allows the model to generate images and responses with a Miyazaki-esque flair—think lush, hand-drawn landscapes and poetic dialogue. Ghibli is helping users view their past, and our favorite memes, in a new lens.
4. Microsoft Introduces Researcher and Analyst AI Agents
Microsoft announced Researcher and Analyst AI agents, which automates data synthesis and insights for professionals.
The new agents will provide deep, expert-level assistance by integrating work emails, meetings, files, and web sources. It’s essentially a mix between OpenAI’s research model and Microsoft 365 Copilot orchestration.
Microsoft will begin rolling out Researcher and Analyst for some Microsoft 365 Copilot users in April.
5. Reve Image 1.0 Outshines Rivals
A new image generator, Reve Image 1.0, burst onto the scene this week, claiming the crown for image generation.
Users are praising how it outperforms established models like DALL-E and Midjourney in detail and realism. Whether it’s photorealistic portraits or surreal dreamscapes, Reve’s debut has artists and developers buzzing about its potential.
Aside from the scores, check out this video created with Reve and see for yourself.