Google announces Gemini 4 Argon AI model, but you can’t use it yet

谷歌发布 Gemini 4 Argon AI 模型，但目前尚无法使用

Google promised Gemini 3.5 Pro in June, but it spent the summer trotting out smaller Flash models. Now, Google is ready to take on the frontier again with Gemini 4 Argon. The company claims this new AI offers industry-leading performance in coding, knowledge work, and cybersecurity, but you aren’t allowed to use it yet.

谷歌曾在六月承诺推出 Gemini 3.5 Pro，但整个夏天却一直在发布较小的 Flash 模型。现在，谷歌准备凭借 Gemini 4 Argon 再次挑战前沿领域。该公司声称，这款新 AI 在编程、知识工作和网络安全方面提供了行业领先的性能，但目前你还无法使用它。

While most of us will have to wait to test Gemini 4, Google says engineers inside the company are already using the new model extensively. Argon reportedly used “fleet-wide telemetry data” to help Google save 300 TiB of memory across its data centers. Meanwhile, Argon agents have been working to migrate C/C++ codebases to Rust across Google, including thousands of lines in the core re2 and libgav1 libraries and more than 800,000 lines in the Fuchsia OS Zircon kernel.

虽然我们大多数人还需要等待才能测试 Gemini 4，但谷歌表示公司内部的工程师已经在大规模使用该模型。据报道，Argon 利用“全舰队遥测数据”帮助谷歌在其数据中心节省了 300 TiB 的内存。与此同时，Argon 智能体一直在致力于将谷歌内部的 C/C++ 代码库迁移至 Rust，其中包括核心 re2 和 libgav1 库中的数千行代码，以及 Fuchsia OS Zircon 内核中超过 80 万行的代码。

Google has also come armed with a raft of benchmarks to back up its claims. On the software engineering DeepSWE v1.1 benchmark, Gemini 4 Argon hits 77.9 percent, which is higher than GPT-6 Astra, Fable 5.1, and Opus 5.5. Google promises similar power across a range of long-horizon tasks, pointing to Argon’s industry-leading score in the economic analysis Vals Index test.

谷歌还准备了一系列基准测试来支持其说法。在软件工程 DeepSWE v1.1 基准测试中，Gemini 4 Argon 达到了 77.9%，高于 GPT-6 Astra、Fable 5.1 和 Opus 5.5。谷歌承诺在各种长周期任务中也能提供类似的能力，并指出了 Argon 在经济分析 Vals Index 测试中取得的行业领先分数。

This model is still in limited testing, but Google has announced API pricing. For a limited time, Argon will offer rates of $2 per million input tokens and $10 per million output tokens, and cached input tokens will be discounted 95 percent. The company has also confirmed that Gemini 4 Argon will support a much higher output limit of 1 million tokens. That’s up from 64,000 tokens in previous Gemini models. Google says this allows users to complete more daunting tasks in a single step.

该模型目前仍处于有限测试阶段，但谷歌已经公布了 API 定价。在限时优惠期间，Argon 的定价为每百万输入 token 2 美元，每百万输出 token 10 美元，且缓存的输入 token 可享受 95% 的折扣。该公司还确认，Gemini 4 Argon 将支持高达 100 万 token 的输出限制，这比之前 Gemini 模型的 64,000 token 有了大幅提升。谷歌表示，这使得用户能够一步完成更艰巨的任务。

The main focus for Gemini 4 right now is cyberdefense. Google says models of this scale call for a phased release, so it’s starting with a small group of trusted testers. Partners in the company’s Fairwind Program can get access to leverage the model’s cybersecurity defense capabilities. Wiz is apparently already using Argon and has used it to uncover a critical vulnerability that could expose personal information in a system used at hospitals around the world. Google claims other frontier models missed this flaw, but it didn’t provide any specifics.

Gemini 4 目前的主要重点是网络防御。谷歌表示，这种规模的模型需要分阶段发布，因此它首先向一小部分受信任的测试人员开放。公司“Fairwind 计划”的合作伙伴可以获得访问权限，以利用该模型的网络安全防御能力。据悉，Wiz 已经在利用 Argon 发现了一个关键漏洞，该漏洞可能导致全球医院所用系统中的个人信息泄露。谷歌声称其他前沿模型未能发现此缺陷，但并未提供具体细节。

There’s a lot of hand-wringing about model misalignment after several high-profile hacking incidents over the summer. Google claims it designed Argon with systems that monitor the model’s chain-of-thought and can stop it in its tracks if the model steps out of bounds. This, Google says, is why reasoning transparency is important in frontier model development.

在今年夏天发生几起备受瞩目的黑客攻击事件后，人们对模型失控（misalignment）感到担忧。谷歌声称，在设计 Argon 时加入了监控模型思维链的系统，如果模型越界，系统可以立即将其停止。谷歌表示，这就是为什么推理透明度在前沿模型开发中至关重要。

Gemini 4 Argon will eventually be available to enterprise and consumer customers, but Google isn’t making any specific timeline promises. We do know that following the cybersecurity tests, general availability will start with paid API users and Google AI Ultra subscribers.

Gemini 4 Argon 最终将面向企业和个人消费者开放，但谷歌并未给出具体的发布时间表。我们已知的是，在网络安全测试之后，该模型将首先向付费 API 用户和 Google AI Ultra 订阅者全面开放。