谷歌终于发布 Gemini 4 Argon

在前沿 AI 竞赛中旁观数月后，谷歌终于推出了备受期待的 Gemini 4 Argon 模型，并称其能够与业内最优秀的模型一较高下。

谷歌表示，Argon 具备“前沿推理能力，并在复杂工作流中表现更出色”，同时减少了仍困扰 AI 聊天机器人的模棱两可、重复和陈词滥调。该公司声称，在涵盖编程、网络安全、科学、金融和法律等领域的一系列基准测试中，Argon 的表现优于 GPT-6 Astra、Claude Fable 5.1 和 Claude Opus 5.5 等竞争模型。

谷歌表示，公司内部已经开始使用 Argon，例如查找和修复网络安全漏洞、将支撑 YouTube 评论区所需的算力减半，以及为 Alphabet 旗下无人机配送公司 Wing 的无人机物理模拟器排查故障。

该模型从即日起向 Gemini 应用和搜索 AI 模式中的高等级 Pro 和 Ultra 订阅者推出，也向其 Agent 开发平台 Antigravity 的 Ultra 订阅者开放。目前尚不清楚 Argon 何时会向 Plus 订阅者或免费用户开放，也不确定是否会开放。

谷歌为 Argon 制定了极具竞争力的价格。该模型的首发优惠价为每百万输入 token 2 美元、每百万输出 token 10 美元，约为 OpenAI 的 GPT-6 Astra 和 Anthropic 的 Fable 5 模型费用的五分之一。即使在 11 月底首发优惠期结束、价格翻倍之后，Argon 的单 token 价格仍将明显低于前沿竞争模型。不过，这并不一定意味着实际使用成本更低，因为不同模型完成相同任务时消耗的 token 数量可能不同。

此次发布补充了谷歌几周前首次预告 Argon时留下的一些信息空白。与近期其他前沿模型的发布方式一致，公开推出之前，Argon 曾有限开放给一组选定的“可信网络防御人员”，以测试潜在风险。

Argon 是谷歌自 2025 年底推出 Gemini 3 以来首次发布的重大前沿模型。在此期间，Anthropic 和 OpenAI 推出的一系列表现亮眼的模型使谷歌远远落后于竞争对手。重大的内部人事和组织调整进一步引发外界对谷歌能否——或是否愿意——重夺地位并在 AI 前沿展开竞争的质疑。

但 Argon 的到来表明，我们还不能就此断定谷歌已退出 AI 竞赛。Meta 刚刚宣布推出自己的前沿模型 Muse Wonder Full，仅一天后 Argon 便告发布。这提醒人们，AI 领域的力量格局可能转变得多么迅速，而前沿竞争远非两家企业之间的较量。

After months sitting on the sidelines of the frontier AI race, Google has finally launched its long-awaited Gemini 4 Argon model, claiming it can compete with the best the industry has to offer.

Google says Argon delivers “frontier reasoning and improved performance across complex workflows,” while cutting down on hedging, repetition, and the cliches that continue to plague AI chatbots. The company claims Argon outperforms rival models like GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 across a range of benchmarks, including coding, cybersecurity, science, finance, and law.

Google says it’s already putting Argon to use internally, such as finding and patching cybersecurity vulnerabilities, halving the computing power needed to power YouTube’s comments section, and debugging the drone physics simulator for Alphabet’s drone delivery company Wing.

The model is rolling out today to Google’s higher-tier Pro and Ultra subscribers in the Gemini app and AI mode in Search, as well to Ultra subscribers on its agentic development platform Antigravity. It is not clear when — or if — Argon will be available to Plus subscribers or free users.

Google is pricing Argon very competitively. The model is launching with an introductory price of $2 per million input tokens and $10 per million output tokens, around a fifth the cost of OpenAI’s GPT-6 Astra and Anthropic’s Fable 5 models. Even after those costs double at the end November — when the introductory period finishes — Argon will still be significantly cheaper than its frontier rivals on a per-token basis, though this doesn’t necessarily translate to lower costs in practice, as models can use different numbers of tokens for the same tasks.

The launch fills in some of the blanks left behind from when Google first teased Argon a few weeks ago. In line with other recent frontier model releases, the public rollout follows a limited release to a select set of “trusted cyber defenders” to be tested for potential risks.

Argon is Google’s first major frontier release since Gemini 3 in late 2025. In the intervening months, impressive models from Anthropic and OpenAI have seen Google trail far behind its rivals. Seismic internal shakeups raised further doubts about the company’s ability — or desire — to reclaim its position and compete at the frontier of AI.

But Argon’s arrival suggests we can’t count Google out of the AI race just yet. Landing just a day after Meta announced its own frontier model, Muse Wonder Full, Argon is a reminder of how quickly the balance of power can shift in AI, and that the frontier is far from a two-horse race.