Gemini 4 Argon rivals GPT-6 Sol, with mixed results

Google has launched Gemini 4 Argon for AI Pro and AI Ultra subscribers, positioning it as an upgrade for complex engineering, scientific research and knowledge work. The company also says it delivers more direct responses and stronger protections against harmful requests, agent overreach and malicious prompt injections. Limited hands-on comparisons with OpenAI’s GPT-6 Sol showed broadly competitive performance, with different strengths depending on the task.

Gemini performed better when asked to build a locally run web app mapping the world’s lighthouses. Both models initially encountered restricted access to online map data and needed another prompt to create their own maps. Gemini’s result offered a more detailed map, animated light ranges and richer location information, while completing the task faster. ChatGPT’s version provided well-organized information and neat lighthouse diagrams, but its mapping was less polished.

ChatGPT had the advantage when summarizing PDFs of 10 recent scientific studies, finishing in 25 seconds versus Gemini’s 55 seconds. Both produced useful summaries, but ChatGPT’s were shorter and easier to understand. For television pitches inspired by Apple TV shows, Gemini responded faster, while ChatGPT’s ideas were judged more nuanced and plausible. These narrow tests suggest neither model is consistently superior across everyday tasks.