2025
Year in Review
RITS @ NYU Shanghai • AI & Emerging Tech
Q1
The DeepSeek Revolution
January 21
DeepSeek R1 Goes Open Source
The open-source release that shocked the industry, producing remarkable code and video outputs from a fully transparent model.
🔓 Open Source
February 2
Janus from DeepSeek
A model completing the cycle—text generation with image understanding AND image creation in one unified system.
🎨 Multimodal
February 25
Claude Sonnet 3.7 with Deep Thinking
Anthropic introduces extended reasoning capabilities, bringing deliberate thinking to AI assistants.
🧠 Reasoning
March 26
GPT-4o Native Image Generation
OpenAI brings native image generation to ChatGPT Plus users, unifying text and visual creation.
🖼️ Generation
March 19
Roblox Cube Launched
Roblox enters the AI-powered 3D experience creation space with their new Cube feature.
🎮 Gaming
March 20
First AI-Generated Peer-Reviewed Paper
Sakana AI achieves a historic milestone—the first actual peer-reviewed scientific paper generated by AI.
📚 Research
Q2
The Agentic Awakening
April 7
ChatGPT Plus Free for Students
OpenAI makes premium AI accessible to college students worldwide, democratizing advanced AI tools.
🎓 Education
April 28
Qwen3 Released
Alibaba drops another major open-source model, intensifying the global AI race.
🔓 Open Source
May 16
OpenAI Codex Launches
OpenAI’s answer to agentic coding—a cloud-based coding agent running on OpenAI servers.
💻 Coding
May 21
Veo 3 Video Generation Beta
Google’s video generation tool goes public with enhanced prompt adherence and physics simulation.
🎬 Video
June 12
Claude Code for Pro Users
Anthropic expands Claude Code access to Pro subscribers, bringing terminal-based AI coding to more developers.
💻 Coding
June 19
MiniMax M1: Million-Token Context
Shanghai-based MiniMax releases the world’s first open-weight model handling one million tokens.
📊 Long Context
June 26
Gemini CLI Announced
Google brings Gemini 2.5 Pro directly to the terminal, enabling native command-line AI development.
🔧 Developer Tools
June 27
FLUX.1 Kontext Open Weights
Black Forest Labs releases 12B parameter open-weights model for advanced image editing.
🎨 Image Editing
Q3
The Frontier Wars
July 16
Kimi K2 from Moonshot AI
Alibaba-backed Moonshot releases a cost-effective open-source model rivaling ChatGPT and Claude.
🔓 Open Source
July 18
ChatGPT Agent Launches
OpenAI bridges research and execution—AI that can browse the web and take real-world actions.
🤖 Agents
July 22
AI Wins IMO Gold
Historic milestone: Both Google’s Gemini and OpenAI models achieve gold medal scores at the International Math Olympiad.
🏅 Milestone
July 28
Qwen3-Coder Unveiled
Alibaba launches its 480B-parameter agentic coding model under Apache 2.0 license.
💻 Coding
August 6
Claude Opus 4.1 Released
Anthropic’s incremental leap brings enhanced coding and agentic capabilities to their flagship model.
🧠 Reasoning
August 8
GPT-5 Official Launch
OpenAI’s most significant upgrade since GPT-4 arrives—available to all users from free tier to enterprise.
🚀 Major Release
August 6
GPT-OSS: OpenAI Goes Open Weight
First open-weight release from OpenAI since GPT-2—120B and 20B models under Apache 2.0.
🔓 Open Source
September 23
DeepSeek V3.1 Terminus
DeepSeek continues pushing boundaries with enhanced NLP, code generation, and multi-lingual support.
🔓 Open Source
Q4
The Next Generation
October 7
Claude Sonnet 4.5 Arrives
Anthropic’s “most aligned frontier model” with major gains in coding, reasoning, and safety.
🧠 Frontier
October 7
Sora 2 + Mobile App
OpenAI unveils next-gen video synthesis with synchronized audio and a dedicated mobile experience.
🎬 Video
October 7
Qwen3-VL Multimodal
Alibaba’s flagship multimodal LLM advances text and vision capabilities for richer understanding.
👁️ Vision
October 15
Intel Crescent Island GPU
160GB LPDDR5x memory GPU designed for massive AI workloads and real-time analytics.
🔧 Hardware
December 15
GPT-5.2: Most Capable Yet
OpenAI’s newest model excels at professional knowledge work, long-context reasoning, and multimodal tasks.
🚀 Frontier
Dominant Themes of 2025
Open Source Renaissance
DeepSeek, Qwen, FLUX, GPT-OSS…
Agentic AI
Claude Code, Codex, ChatGPT Agent, Manus
Video Generation
Veo 3, Sora 2, Wan2.2, HunyuanVideo
Reasoning Models
GPT-5, Claude 4.5, DeepSeek R1
AI for Education
Free student access, Claude Academy
China’s AI Rise
DeepSeek, Qwen, MiniMax, Moonshot
Leave a Reply