
Mercury 2, a new reasoning LLM from Inception Labs, employs a diffusion architecture that creates and refines tokens in parallel, achieving roughly 1,000 tokens per second. In live tests, the model wrote a complete checkers game and a full chess engine within seconds, demonstrating rapid code generation and responsive follow‑up edits. Compared with Claude Haiku, Mercury 2 completed the same tasks significantly faster while preserving strong reasoning capabilities. The speed and reasoning mix positions it for AI agents, coding assistants, voice interfaces, search, and customer‑service applications.

The video pits the newest AI image generators—Google’s Nano Banana Pro, OpenAI’s ChatGPT image model, Flux 2 Pro, and Midjourney—against each other in a systematic, 15‑prompt showdown. The creator walks through each platform’s interface, feeds identical prompts, and evaluates the outputs on realism, text...