Google Gemini 4 Argon closes the gap with OpenAI and Anthropic but doesn't take a clear lead
decoder ·
Google has unveiled Gemini 4 Argon, its new flagship AI model designed to close the competitive gap with rivals OpenAI and Anthropic. While it does not claim a clear overall lead, the model matches top-tier performance on key benchmarks and introduces an industry-first one million output token limit for complex reasoning. Independent assessments from Artificial Analysis show Argon tying with GPT-6 Astra in intelligence scores but consuming significantly more tokens per task. Despite higher token usage, Google’s introductory pricing makes it cost-effective, although regular rates may reduce this advantage. The rollout is phased, beginning with trusted cyber defenders before expanding to developers and consumers. Early tests highlight strong improvements in agentic tasks and low hallucination rates compared to competitors. However, Google’s surrounding software ecosystem still lags behind rival offerings like Claude Cowork.
decoder