AI Developments in 2025: Innovations and Trends

AI has taken centre stage in 2025, featuring remarkable launches, new generation AI models, and innovative technologies that have integrated deeply into daily life. Among various tech leaders, Google made significant strides this year with the introduction of the Gemini 3 model, highlighting considerable enhancements in reasoning skills, multimodal comprehension, and real-time productivity across the Gemini application and its associated ecosystem.

Meanwhile, OpenAI also positioned itself competitively with the release of GPT-5 and GPT-5.2, which provided quicker responses, enhanced contextual understanding, and a range of AI-driven tools. Besides these primary updates, the year further witnessed notable progress in AI-powered creative tools, on-device intelligence, and additional capabilities.

Top AI News of 2025

Google Gemini 3.0

This year, Google launched its most advanced and intelligent AI model to date, the Gemini 3. The latest AI models feature enhanced reasoning capabilities, improved multimodal understanding, and tighter integration across Google Search, Workspace, and Android. The excitement surrounding Gemini 3 stems from its advanced agentic features that facilitate planning of multi-step tasks and more fluid interactions across apps, acting as a collaborative partner for users.

Perplexity Comet Browser

In September, Perplexity introduced its AI-native web browser named Comet. Built on the Chromium framework, this browser utilises autonomous agentic AI to undertake tasks on behalf of the user. Comet can manage browsing sessions, navigate websites, complete forms, and carry out online transactions.

OpenAI Atlas (AI Browser)

In response to Comet and Google’s AI Mode, OpenAI launched its own AI-driven browser called Atlas. This interactive and agentic AI browser is capable of executing multi-step actions, online shopping, or thorough research for users. The “Memory” function enhances personalisation by allowing users to revisit previous searches with a single prompt.

Sora 2 Text-to-Video Generation

Sora 2, the second-generation video and audio model from OpenAI, gained significant traction upon its release this year. The company rolled out several new features, including Cameos, Remix culture, and a social media-like AI-native Feed for sharing AI-generated videos. Although currently available only in the US and Canada, wider accessibility is anticipated next year.

OpenAI’s ChatGPT 5.2

Earlier in the year, OpenAI introduced its flagship AI model, GPT-5. Subsequently, December 2025 saw the launch of an upgraded version labelled GPT-5.2, showcasing capabilities such as human-expert reasoning, improved visual recognition, and more. This model is also optimised for agentic workflows, creating direct competition with Google’s Gemini 3 model.

Adobe Max AI Suite

Finally, Adobe made pivotal announcements in the AI sector this year, beginning with the introduction of the Firefly Image Model 5 for AI-generated images. The company also unveiled a new video model along with tools like Generate Soundtrack, Generate Speech, and the Firefly Video Editor. Additionally, Adobe presented its conversational AI assistant, Photoshop Express AI Assistant, designed to carry out edits on behalf of users.