Gemini 4 Argon 正式发布：API 定价与内部落地案例

Martin Varsavsky @martinvars原推发布于 2026.10.14

在 X 上打开原推

Google just released Gemini 4 Argon to the world. Today developers, enterprises and Pro and Ultra subscribers get Google DeepMind's new frontier model. Google says it ranks first on the Vals Index, which scores models on finance, coding, legal and tax work. The introductory API price is two dollars per million input tokens and ten per million output, roughly a fifth of what OpenAI and Anthropic charge for their top models. After November 30 that doubles, and it is still cheap by frontier standards. What already happened inside Google matters more than the benchmarks. Argon helped the Chrome team finish an architectural refactor across hundreds of files in a codebase of about forty million lines, months of work done in weeks. It cut in-memory session prep time by forty percent in their open-source Agent Development Kit. Researchers used it to make the quantum chemistry package xTB up to 2.3 times faster. At Isomorphic Labs it turns weeks of reading patents and papers into hours. They are also giving ten thousand scientists six months of the highest Antigravity tier for free. The API carries a one million token unified context window. When frontier intelligence gets this cheap, the constraint stops being the model. It becomes how fast you put it to work.