5 upgrades Gemini needs to beat ChatGPT (www.techradar.com)

TechRadar lays out five concrete upgrades Google’s upcoming Gemini 3 needs to outcompete ChatGPT: reliable long-term memory, faster but smarter responses, better intent understanding, deeper multimodal perception, and true task-completion abilities. Despite iterative gains from Gemini 1.5 through 2.5, users still default to ChatGPT because it feels more consistent and useful day-to-day. Gemini 3 is framed as Google’s best chance to move from impressive demos to a genuinely dependable assistant that people open first. Technically, that means shipping robust stateful memory and context-tracking across long chats and documents; adaptive inference (model-routing or “thinking” modes) to balance latency and depth the way ChatGPT 5 does; explicit intent-disambiguation (follow-up prompts rather than confident misinterpretations); richer multimodal reasoning beyond surface-level image/video captions (form analysis, procedural guidance); and agentic, multi-step task execution with error recovery, API integration and planning. If Google ties these capabilities together — leveraging its data and ecosystem — Gemini could shift from a smart search-like bot to a proactive, reliable assistant that outperforms current conversational norms.