Change
93392ae
93392ae03196a5f44bf00935cbf885dfc3ae6d97 · commit on GitHub
gcloud-blog-feed: changed (493068 bytes, HTTP 200)
raw/gcloud-blog-feed/response.xml modified
- Source
- gcloud-blog-feed
- Lines added
- +252
- Lines removed
- -696
- Stored bytes at this commit
- 493,068
- Timestamp
- observed
- Raw artifact at this commit
- raw/gcloud-blog-feed/response.xml
Recorded headers
| observed_at | 2026-11-04T06:22:02.159Z |
|---|---|
| origin_date | null |
| status | 200 |
| final URL | https://cloudblog.withgoogle.com/products/ai-machine-learning/rss/ |
| etag | null |
| last-modified | null |
| date | Wed, 04 Nov 2026 06:22:02 GMT |
| age | null |
| cache-control | public, max-age=600, s-maxage=600 |
| cf-cache-status | null |
| content-encoding | gzip |
| content-length | 131685 |
@
@@ -1,5 +1,248 @@<?xml version="1.0" encoding="utf-8"?>-
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:media="http://search.yahoo.com/mrss/"><channel><title>AI & Machine Learning</title><link>https://cloud.google.com/blog/products/a…+
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:media="http://search.yahoo.com/mrss/"><channel><title>AI & Machine Learning</title><link>https://cloud.google.com/blog/products/a…+
<p><span style="vertical-align: baseline;">To solve these long-standing challenges, fintech innovator </span><a href="https://www.nammu21.com/" rel="noopener" target="_blank"><span style="text-decoration: underline; vertical-align: baseline;">Nammu21</span></a&…+
<h2><span style="vertical-align: baseline;">Digitizing the loan market</span></h2>+
<p><span style="vertical-align: baseline;">While loan markets provide billions in capital to businesses worldwide, their underlying infrastructure has not kept pace with the speed of digital finance. This landscape requires continuous coordination across facility agents, borrowers, and d…+
<p><span style="vertical-align: baseline;">In particular, relying on fragmented, manual interpretation of legal text often results in significant operational drag and data silos. According to the </span><a href="https://blog.lma.eu.com/challenging-the-timeline-issue-no-4/" rel="…+
<p><span style="vertical-align: baseline;">Nammu21 transforms bespoke credit documents into programmable assets at scale by rendering each facility's governing terms as a Digital Credit Instrument (DCI). The DCI maps every term — including conditions, consents, eligibility criteria, and …+
<p><span style="vertical-align: baseline;">To instantiate these instruments, the proof-of-concept utilizes the secure, high-performance execution environment of GCUL, an institutional-grade, permissioned distributed ledger that helps automate complex financial workflows to deliver near-i…+
<h2><span style="vertical-align: baseline;">A purpose-built stack for digital credit</span></h2>+
<p><span style="vertical-align: baseline;">Combined with Nammu21’s data intelligence, GCUL provides the foundational infrastructure for converting legal document logic into live, programmable contracts. Designed to handle the privacy, scale, and multi-party coordination required by insti…+
<p><span style="vertical-align: baseline;">Programmable contracts on GCUL are written in </span><code style="vertical-align: baseline;">gculpy</code><span style="vertical-align: baseline;">, GCUL's domain-optimized programming language. As a strict, statically-typ…+
<p><span style="vertical-align: baseline;">For regulated financial institutions, </span><code style="vertical-align: baseline;">gculpy</code><span style="vertical-align: baseline;"> offers distinct operational advantages over other ledger programming models. The l…+
<p><span style="vertical-align: baseline;">GCUL also provides robust, built-in permissioning, requiring explicit user authorization before a contract can store data on their behalf. This solves the "unsolicited asset" problem — common on public networks — where unwanted tokens can be sen…+
<p><span style="vertical-align: baseline;">Nammu21 serves a dual role in this architecture. First, the platform renders the DCI programmatically from the original legal agreement without manual document reading to establish the terms. Next, it derives </span><code style="vertica…+
<h2><span style="vertical-align: baseline;">Technical architecture: The loan lifecycle on GCUL</span></h2>+
<p><span style="vertical-align: baseline;">Nammu21’s loan implementation on GCUL demonstrates the value of creating a single, immutable source of truth. The proof-of-concept modeled a $1.945 billion Term Loan B — one segment of a broader multi-facility transaction — where the agent bank,…+
<p><span style="vertical-align: baseline;">The lifecycle begins when an agent bank selects a specific credit agreement transformed by Nammu21. The platform instantly provides the structured terms from the document, which are used to instantiate a unique </span><code style="verti…+
<div class="block-image_full_width">+
+
+
+
+
+
+
+
<div class="article-module h-c-page">+
<div class="h-c-grid">+
+
+
<figure class="article-image--large+
+
+
h-c-grid__col+
h-c-grid__col--6 h-c-grid__col--offset-3+
+
+
"+
>+
+
+
+
+
<img+
src="https://storage.googleapis.com/gweb-cloudblog-publish/images/1-facility-tokenization.max-1000x1000.png"+
+
alt="1-facility-tokenization">+
+
</a>+
+
</figure>+
+
+
</div>+
</div>+
+
+
+
+
+
</div>+
<div class="block-paragraph_advanced"><p><span style="vertical-align: baseline;">Next, the agent bank initiates the onboarding of the borrower and lenders. This multilateral action requires that each participant review the request and then provide explicit consent by approving the …+
<p><span style="vertical-align: baseline;">Subsequent lifecycle events, such as secondary market transfers or repayments, are also managed as multi-step, automated workflows governed by a consent-driven model. For example, a transfer of a loan position requires new lender, borrower, and …+
<div class="block-image_full_width">+
+
+
+
+
+
+
+
<div class="article-module h-c-page">+
<div class="h-c-grid">+
+
+
<figure class="article-image--large+
+
+
h-c-grid__col+
h-c-grid__col--6 h-c-grid__col--offset-3+
+
+
"+
>+
+
+
+
+
<img+
src="https://storage.googleapis.com/gweb-cloudblog-publish/images/2-assignment-transfer-legend.max-1000x1000.png"+
+
alt="2-assignment-transfer-legend">+
+
</a>+
+
</figure>+
+
+
</div>+
</div>+
+
+
+
+
+
</div>+
<div class="block-paragraph_advanced"><p><span style="vertical-align: baseline;">Transacting with familiar, regulated money on a programmable ledger provides an architecture purpose-built for institutional finance, grounded in three core pillars:</span></p>+
<p><span style="vertical-align: baseline;">1. </span><strong style="vertical-align: baseline;">Compliance via gated execution: </strong><span style="vertical-align: baseline;">Nammu21’s </span><code style="vertical-align: baseline;">gculpy</code>…+
<p><span style="vertical-align: baseline;">2. </span><strong style="vertical-align: baseline;">Role-based privacy:</strong><span style="vertical-align: baseline;"> GCUL enforces strict, role-based access, ensuring that trading strategies, positions, and sensitive …+
<p><span style="vertical-align: baseline;">3. </span><strong style="vertical-align: baseline;">Settlement finality: </strong><span style="vertical-align: baseline;">Every on-ledger action completes near-instantly and anchors on canonical data derived from the orig…+
<h2><span style="vertical-align: baseline;">The business impact of moving on-ledger</span></h2>+
<p><span style="vertical-align: baseline;">This proof-of-concept illustrates a structural shift in how credit markets can operate. When each facility's governing terms are rendered as a DCI and executed on GCUL, the operational mechanics of lending markets change fundamentally, rather th…+
<p><span style="vertical-align: baseline;">Values can be reconciled bilaterally, but governance logic cannot – and the market’s current friction stems directly from each party's independent interpretation of the rules. Resolving this issue on-ledger is what enables settlement times to co…+
<p><span style="vertical-align: baseline;">This same architecture can be applied wherever credit terms are bespoke and manually interpreted. By freeing up capital that would otherwise be trapped in lengthy settlement cycles, new liquidity is unlocked and a more dynamic and efficient seco…+
<hr/>+
<p><sup><span style="font-style: italic; vertical-align: baseline;">The team would also like to thank Someera Khokhar, Jordan Shelley, and Saskia A. Aalders at Nammu21 and Googlers Ishan Singh, Kathryn Stavish, Tom Chudy, and Stathis Onasoglou for their contributions.</span>&…+
<hr/>+
<p><span style="vertical-align: baseline;">Today, the </span><a href="http://aiinfra.google" rel="noopener" target="_blank"><span style="text-decoration: underline; vertical-align: baseline;">Google AI &amp; Infrastructure</span></a><span style="verti…+
<p><span style="vertical-align: baseline;">To help the research community measure and improve how agents handle physical engineering work, we are releasing 29 public tasks, their grading tools, and an open competition leaderboard at </span><a href="http://ple-bench.com" rel="noo…+
<p><span style="vertical-align: baseline;">Frontier labs and the broader research community use model examinations to test AI’s progress in completing tasks in a specific discipline. </span><a href="https://lastexam.ai/" rel="noopener" target="_blank"><span style="text-dec…+
<h3><span style="vertical-align: baseline;">Why textbook benchmarks fall short for physical engineering</span></h3>+
<p><span style="vertical-align: baseline;">Frontier language models now score above 90% on many academic benchmarks. However, those benchmarks mostly test textbook-style questions, single tool calls, or small, self-contained design problems. For AI developers and researchers building age…+
<h3><span style="vertical-align: baseline;">Inside PLE: Real-world tasks built with practicing engineers</span></h3>+
<p><span style="vertical-align: baseline;">PLE closes this evaluation gap with 85 long-horizon tasks across four core disciplines: </span><strong style="vertical-align: baseline;">mechanical</strong><span style="vertical-align: baseline;"> (30 tasks), </span>…+
<div class="block-image_full_width">+
+
+
+
+
+
+
+
<div class="article-module h-c-page">+
<div class="h-c-grid">+
+
+
<figure class="article-image--large+
+
+
h-c-grid__col+
h-c-grid__col--6 h-c-grid__col--offset-3+
+
+
"+
>+
+
+
+
+
<img+
src="https://storage.googleapis.com/gweb-cloudblog-publish/images/PLE_Kaggle_-_ple_disciplines_description.max-1000x1000.png"+
+
alt="PLE Kaggle - ple_disciplines_description">+
+
</a>+
+
</figure>+
+
+
</div>+
</div>+
+
+
+
+
+
</div>+
<div class="block-paragraph_advanced"><p><span style="vertical-align: baseline;">Instead of answering a single prompt, an agent works inside a containerized project folder — built on</span><a href="https://github.com/Mercor-Intelligence/archipelago" rel="noopener" target="…+
<p><span style="vertical-align: baseline;">In collaboration with experts from Mercor and licensed practicing engineers from industry, we built PLE using two methods:</span></p>+
<ul>+
<li aria-level="1" style="list-style-type: disc; vertical-align: baseline;">+
<p role="presentation"><span style="vertical-align: baseline;">Expert-authored tasks (44 tasks): Practicing engineers created the project folders, prompts, and grading rubrics based on real professional engineering tasks.</span></p>+
</li>+
<li aria-level="1" style="list-style-type: disc; vertical-align: baseline;">+
<p role="presentation"><span style="vertical-align: baseline;">Co-authored tasks (41 tasks): Practicing engineers designed the initial task and an agent pipeline assembled the task's environment and grader from licensable engineering standards. Engineers then validated and increased the …+
</li>+
</ul>+
<h3><span style="vertical-align: baseline;">All frontier models score below 50%</span></h3>+
<p><span style="vertical-align: baseline;">We tested 14 frontier models on PLE using seven publicly available agent harnesses (Antigravity, Claude Code, Codex, Hermes, OpenCode, OpenHands, and Pi). Because PLE evaluates both code compliance and cost efficiency, agents only earn points if…+
<ul>+
<li aria-level="1" style="list-style-type: disc; vertical-align: baseline;">+
<p role="presentation"><span style="vertical-align: baseline;">High headroom: All frontier models score below 50% on our private test set, with the highest-performing model on its native harness reaching a mean score of 48.3%.</span></p>+
</li>+
<li aria-level="1" style="list-style-type: disc; vertical-align: baseline;">+
<p role="presentation"><span style="vertical-align: baseline;">Hard safety gates: While the top-scoring agent configuration satisfied 91% of individual constraints on co-authored tasks, only 41.3% of its deliverables passed every hard safety and feasibility check required.</span>&l…+
</li>+
<li aria-level="1" style="list-style-type: disc; vertical-align: baseline;">+
<p role="presentation"><span style="vertical-align: baseline;">Model vs. harness impact: Choice of model drives performance far more than the harness (up to a 50-point spread across models, versus a median 3-point spread across harnesses for the same model), though harnesses significantl…+
</li>+
</ul></div>+
<div class="block-image_full_width">+
+
+
+
+
+
+
+
<div class="article-module h-c-page">+
<div class="h-c-grid">+
+
+
<figure class="article-image--large+
+
+
h-c-grid__col+
h-c-grid__col--6 h-c-grid__col--offset-3+
+
+
"+
>+
+
+
+
+
<img+
src="https://storage.googleapis.com/gweb-cloudblog-publish/images/ple_frontier_private_x.max-1000x1000.png"+
+
alt="ple_frontier_private_x">+
+
</a>+
+
</figure>+
+
+
</div>+
</div>+
+
+
+
+
+
</div>+
<div class="block-paragraph_advanced"><h3><strong style="vertical-align: baseline;">Test your agent and join the open competition</strong></h3>+
<p><span style="vertical-align: baseline;">Evaluating AI agents on realistic project files, strict safety rules, and material costs is a practical step toward building helpful tools for physical engineers. Here is how you can get started with PLE today:</span></p>+
<ul>+
<li aria-level="1" style="list-style-type: disc; vertical-align: baseline;">+
<p role="presentation"><span style="vertical-align: baseline;">Check out the full technical breakdown in our </span><a href="https://arxiv.org/pdf/2610.30537" rel="noopener" target="_blank"><span style="text-decoration: underline; vertical-align: baseline;">PLE paper on…+
</li>+
<li aria-level="1" style="list-style-type: disc; vertical-align: baseline;">+
<p role="presentation"><span style="vertical-align: baseline;">Download the 29 public tasks, grading scripts, and evaluation container from our </span><a href="https://github.com/google-research/ple" rel="noopener" target="_blank"><span style="text-decoration: underline; v…+
</li>+
<li aria-level="1" style="list-style-type: disc; vertical-align: baseline;">+
<p role="presentation"><span style="vertical-align: baseline;">Bring your own agent setup and submit it for evaluation on our private test set at </span><a href="http://ple-bench.com" rel="noopener" target="_blank"><span style="text-decoration: underline; vertical-align: b…+
</li>+
</ul></div></description><pubDate>Tue, 03 Nov 2026 08:00:00 +0000</pubDate><guid>https://cloud.google.com/blog/products/ai-machine-learning/introducing-physical-engineerings-last-exam-for-agents/</guid><category>AI & Machine Learning</category><og xmlns:og="http://ogp.me/ns#"><type>a…<div class="block-paragraph_advanced"><p><span style="vertical-align: baseline;">This October, we took massive leaps forward in delivering flexible, open, and powerful enterprise AI solutions. From bringing Gemini 4 Argon to Google Cloud to expanding choice and flexibility, we’re g…<p><span style="vertical-align: baseline;">Here’s a quick look at the major milestones, expert insights, and practical developer guides to read through and save. </span></p><h3><strong style="vertical-align: baseline;">Top announcements </strong></h3>@
@@ -44,7 +287,7 @@<p><span style="vertical-align: baseline;">Stay tuned for monthly updates on Google Cloud’s AI announcements, news, and best practices. For a deeper dive into the latest from Google Cloud customers, read our monthly recap, </span><a href="https://cloud.google.com/blog/topics/cus…<div class="block-aside"><dl> <dt>aside_block</dt>-
<dd>&lt;ListValue: [StructValue([(&#x27;title&#x27;, &#x27;$300 in free credit to try Google Cloud AI and ML&#x27;), (&#x27;body&#x27;, &lt;wagtail.rich_text.RichText object at 0x7fc7e8e62c10&gt;), (&#x27;btn_text&#x27;, &#x27;Start building …+
<dd>&lt;ListValue: [StructValue([(&#x27;title&#x27;, &#x27;$300 in free credit to try Google Cloud AI and ML&#x27;), (&#x27;body&#x27;, &lt;wagtail.rich_text.RichText object at 0x7f4ed45903d0&gt;), (&#x27;btn_text&#x27;, &#x27;Start building …</dl></div><div class="block-paragraph_advanced"><hr/><h2 style="text-align: center;">September</h2>@
@@ -699,7 +942,7 @@ </section></div>-
</div></description><pubDate>Mon, 02 Nov 2026 16:00:00 +0000</pubDate><guid>https://cloud.google.com/blog/products/ai-machine-learning/what-google-cloud-announced-in-ai-this-month/</guid><category>Google Cloud</category><category>AI & Machine Learning</category><media:content height="540" …+
</div></description><pubDate>Mon, 02 Nov 2026 16:00:00 +0000</pubDate><guid>https://cloud.google.com/blog/products/ai-machine-learning/what-google-cloud-announced-in-ai-this-month/</guid><category>Google Cloud</category><category>AI & Machine Learning</category><media:content height="540" …<p><span style="vertical-align: baseline;">With this release, we’re delivering ways to protect your workloads from runaway costs without risking resource deletion. </span><strong style="vertical-align: baseline;">More importantly, we’re introducing three major updates, direct…<h3><span style="vertical-align: baseline;">How the new spend caps can help your business </span></h3><p><span style="vertical-align: baseline;">We have significantly enhanced the usability and power of spend caps in three ways: </span></p>@
@@ -1434,15 +1677,15 @@</div></div></div>-
<p><span style="vertical-align: baseline;">No matter how you choose to deploy, you can access Anthropic's latest frontier models on Google Cloud, including </span><a href="https://console.cloud.google.com/agent-platform/publishers/anthropic/model-garden/claude-fable-5-1"><…+
<p><span style="vertical-align: baseline;">No matter how you choose to deploy, you can access Anthropic's latest frontier models on Google Cloud, including </span><a href="https://console.cloud.google.com/agent-platform/publishers/anthropic/model-garden/claude-fable-5-1"><…<div class="block-code"><dl> <dt>code_block</dt>-
<dd>&lt;ListValue: [StructValue([(&#x27;code&#x27;, &#x27;---\r\nname: test-triager\r\ndescription: Runs unit tests, classifies failures, and returns a concise summary of failing assertions.\r\nmodel: flash\r\ntools:\r\n - read\r\n - edit\r\n---\r\n\r\nRun the requested t…+
<dd>&lt;ListValue: [StructValue([(&#x27;code&#x27;, &#x27;---\r\nname: test-triager\r\ndescription: Runs unit tests, classifies failures, and returns a concise summary of failing assertions.\r\nmodel: flash\r\ntools:\r\n - read\r\n - edit\r\n---\r\n\r\nRun the requested t…</dl></div><div class="block-paragraph"><p data-block-key="5p921">And a cross-model code reviewer subagent in <code>.agents/agents/gemini-judge.md</code>:</p></div><div class="block-code"><dl> <dt>code_block</dt>-
<dd>&lt;ListValue: [StructValue([(&#x27;code&#x27;, &quot;---\r\nname:cross-model-judge\r\ndescription: Independent LLM-as-a-judge verifier that audits diffs and edge cases using Gemini Pro.\r\nmodel: pro\r\ntools:\r\n - read\r\n---\r\n\r\nReview the current workspace diff…+
<dd>&lt;ListValue: [StructValue([(&#x27;code&#x27;, &quot;---\r\nname:cross-model-judge\r\ndescription: Independent LLM-as-a-judge verifier that audits diffs and edge cases using Gemini Pro.\r\nmodel: pro\r\ntools:\r\n - read\r\n---\r\n\r\nReview the current workspace diff…</dl></div><div class="block-paragraph"><p data-block-key="5p921">With this setup, you can delegate rapid test-and-classify loops to <code>test-triager</code> and trigger <code>cross-model-judge</code> to validate the final diff, combining the strengths of each model in a si…<div class="block-video">@
@@ -2631,7 +2874,7 @@<p><code style="vertical-align: baseline;">AI.COUNT_TOKENS</code><span style="vertical-align: baseline;"> returns the number of tokens a model's tokenizer produces for a text string, without making a billable generation call. Run it over a column to measure your input data:&l…<div class="block-code"><dl> <dt>code_block</dt>-
<dd>&lt;ListValue: [StructValue([(&#x27;code&#x27;, &quot;-- Count the input text tokens for each product review\r\nSELECT\r\n SUM(\r\n AI.COUNT_TOKENS(\r\n CONCAT(&#x27;Classify the sentiment of this review as positive, negative, or neutral: &#x27;, review…+
<dd>&lt;ListValue: [StructValue([(&#x27;code&#x27;, &quot;-- Count the input text tokens for each product review\r\nSELECT\r\n SUM(\r\n AI.COUNT_TOKENS(\r\n CONCAT(&#x27;Classify the sentiment of this review as positive, negative, or neutral: &#x27;, review…</dl></div><div class="block-paragraph_advanced"><p><span style="vertical-align: baseline;">The count is a lower bound on input tokens. It doesn't include output or thinking tokens, the instructions each AI function adds around your input, or non-text inputs such as images. Use it to:</spa…<ul>@
@@ -2658,7 +2901,7 @@<p><span style="vertical-align: baseline;">If a small sample run shows an average </span><strong style="vertical-align: baseline;">100:1 input-to-output token ratio</strong><span style="vertical-align: baseline;">, you can solve for a daily budget:</span><…<div class="block-code"><dl> <dt>code_block</dt>-
<dd>&lt;ListValue: [StructValue([(&#x27;code&#x27;, &#x27;Cost: 0.125 × I + 0.75 × O = 100 (I, O in millions of tokens)\r\nRatio: I = 100 × O&#x27;), (&#x27;language&#x27;, &#x27;&#x27;), (&#x27;caption&#x27;, &lt;wagtail.rich_text.Ric…+
<dd>&lt;ListValue: [StructValue([(&#x27;code&#x27;, &#x27;Cost: 0.125 × I + 0.75 × O = 100 (I, O in millions of tokens)\r\nRatio: I = 100 × O&#x27;), (&#x27;language&#x27;, &#x27;&#x27;), (&#x27;caption&#x27;, &lt;wagtail.rich_text.Ric…</dl></div><div class="block-paragraph_advanced"><p><span style="vertical-align: baseline;">Solving gives about </span><strong style="vertical-align: baseline;">754 million input tokens</strong><span style="vertical-align: baseline;"> and </span><strong style=…<h3><span style="vertical-align: baseline;">Best practices for managing AI cost</span></h3>@
@@ -2690,691 +2933,4 @@</li></ol><p><span style="vertical-align: baseline;">For the full details, see the documentation on </span><a href="https://docs.cloud.google.com/bigquery/docs/control-genai-costs"><span style="text-decoration: underline; vertical-align: baseline;">controlling generative AI costs…-
<p><span style="vertical-align: baseline;">With token-based guardrails in place, your teams can use generative AI in BigQuery with the cost predictability your organization needs.</span></p></div></description><pubDate>Wed, 14 Oct 2026 16:00:00 +0000</pubDate><guid>http…-
<p><span style="vertical-align: baseline;">To enable a new frontier of work, we are making our new frontier model, </span><a href="https://deepmind.google/models/gemini/" rel="noopener" target="_blank"><span style="text-decoration: underline; vertical-align: baseline;">…-
<p><span style="vertical-align: baseline;">We trained Argon to work with both depth and precision, which allows it to maintain context across long, multi-step tasks which are reinforced with strong domain expertise along the way. Whether your team is running complex software builds, hand…-
<div class="block-image_full_width">-
-
-
-
-
-
-
-
<div class="article-module h-c-page">-
<div class="h-c-grid">-
-
-
<figure class="article-image--large-
-
-
h-c-grid__col-
h-c-grid__col--6 h-c-grid__col--offset-3-
-
-
"-
>-
-
-
-
-
<img-
src="https://storage.googleapis.com/gweb-cloudblog-publish/images/Gemini_4_Logo_Wall_3.max-1000x1000.png"-
-
alt="Logo Wall Argon V3">-
-
</a>-
-
<figcaption class="article-image__caption "><p data-block-key="l3qrs">Customers and partners using Gemini 4 Argon</p></figcaption>-
-
</figure>-
-
-
</div>-
</div>-
-
-
-
-
-
</div>-
<div class="block-paragraph_advanced"><h3><span style="vertical-align: baseline;">Four ways Argon changes how your business works</span></h3>-
<h4><strong style="vertical-align: baseline;">1. Hand off your hardest work</strong></h4>-
<p><span style="vertical-align: baseline;">Your teams already use Gemini for business to answer questions and draft content. With Argon, they can hand off the deep, multi-step work that used to take hours or days, and get back finished work they can use.</span></p>-
<p><span style="vertical-align: baseline;">Argon reads long documents that mix text, tables, charts and images, and turns them into one clear output grounded in those sources. It can pull findings from many sources into a single structured answer, analyze data over time, and write strong…-
<p><span style="vertical-align: baseline;">Customers who tested Argon tell us it excels at turning huge, unstructured inputs into cohesive outputs. </span></p>-
<p><span style="vertical-align: baseline;">"As software development becomes more agentic, teams need models that can support long-running, multi-step work across all stages of the development lifecycle. Gemini 4 Argon is designed to help organizations bring AI into development, security,…-
<h4><strong style="vertical-align: baseline;">2. Precise, verifiable results on the work that drives your business</strong></h4>-
<p><span style="vertical-align: baseline;">In finance, legal, tax, and other regulated work, calculations and data extraction need to be precise, repeatable, and verifiable. </span></p>-
<p><span style="vertical-align: baseline;">For financial quantitative work, Argon can help automate multi-step ledger reconciliations, spreadsheet formulas, and variance analyses. It delivers high calculation integrity, citation fidelity, and formula accuracy across quantitative models. …-
<p><span style="vertical-align: baseline;">In legal research, Argon can help legal teams review case law and regulatory filings to pull out verified citations and key precedents. It leads on domain-specific evaluations like Harvey's Legal Agent Benchmark for legal research and drafting, …-
<p><span style="vertical-align: baseline;">"Gemini 4 Argon shows how much progress Google has made on agentic legal reasoning. We saw substantial improvements over Gemini 3.1 Pro across our legal agent evaluations, particularly in corporate M&amp;A, data privacy and cybersecurity, an…-
<p><span style="vertical-align: baseline;">“We collaborated with Google on NowAI-Bench to help shape Gemini 4 for key enterprise work. In early access, it delivered roughly 20% improved task success than Gemini 3.8 Flash, with standout gains in HR and customer service. On the ServiceNow …-
<h4><strong style="vertical-align: baseline;">3. Build and ship products faster</strong></h4>-
<p><span style="vertical-align: baseline;">Argon is built to handle complex, multi-step projects from start to finish. Our customers consistently ask for intelligence that can break down goals, track progress, and update multiple files at the same time.</span></p>-
<p><span style="vertical-align: baseline;">"Gemini 4 Argon handles complex tasks reliably, from 3D games to rich animations. In Figma Make, it helps teams turn their most ambitious, multi-step ideas into impressive prototypes.” <strong>– Loredana Crisan, Chief Design Officer, Figma…-
<p><span style="vertical-align: baseline;">To see how well Argon handles this, we put it to work on our own tools. The team behind our open-source Agent Development Kit (ADK) set out to reduce framework overhead and improve code performance while maintaining full compatibility with exist…-
<p><span style="vertical-align: baseline;">Across Google, engineering teams using Argon in Google Antigravity have increased agentic code submissions by more than 35% while cutting code rollbacks by nearly a third. The Chrome team used Argon to finish an architectural refactor across hun…-
<p><span style="vertical-align: baseline;">For more advanced algorithms, our code optimization and discovery agent </span><a href="https://cloud.google.com/blog/products/ai-machine-learning/alphaevolve-is-available-for-everyone"><span style="text-decoration: underline; ver…-
<p><span style="vertical-align: baseline;">To learn how to best use Argon, check out the developer's </span><a href="https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-4-argon"><span style="text-decoration: underline; vertical-align: baseli…-
<h4><strong style="vertical-align: baseline;">4. Stronger defense for your code and business </strong></h4>-
<p><span style="vertical-align: baseline;">While Argon has cyber mitigations that block offensive tasks such as pentesting and exploitation,</span><strong style="vertical-align: baseline;"> </strong><span style="vertical-align: baseline;">you can still take full a…-
<p><span style="vertical-align: baseline;">"We are integrating Gemini 4 Argon into MarketingX for automated codebase security audits. In our evaluations, the model successfully identified 100% of top-priority vulnerabilities with 0% fabricated vulnerabilities across verified claims. This…-
<p><span style="vertical-align: baseline;">Security also means being able to trust Argon inside your own environment. </span><span style="vertical-align: baseline;">Argon is our most resilient model yet against indirect prompt injections</span><span style="vertical-a…-
<p><span style="vertical-align: baseline;">We are also releasing Argon without cyber guardrails to a select set of trusted defenders, which allows for testing a full breadth of cybersecurity capabilities. Trusted defenders are vetted through our </span><a href="https://deepmind.…-
<h3><span style="vertical-align: baseline;">Enterprise customers and partners in our early access program are already seeing strong wins with Argon.</span></h3></div>-
<div class="block-video">-
-
-
-
<div class="article-module article-video ">-
<figure>-
<a class="h-c-video h-c-video--marquee"-
href="https://youtube.com/watch?v=opNsprLPCKc"-
data-glue-modal-trigger="uni-modal-opNsprLPCKc-"-
data-glue-modal-disabled-on-mobile="true">-
-
-
-
-
<div class="article-video__aspect-image"-
style="background-image: url(https://storage.googleapis.com/gweb-cloudblog-publish/images/Teaser_2.2_approved.max-1000x1000.png);">-
<span class="h-u-visually-hidden">Business leaders discuss how they are testing and building with Gemini 4 Argon.</span>-
</div>Diff display stops at 400 lines. The line counts above are from the whole diff. 78 lines shown here cut at 300 characters. The raw artifact at this commit is linked above.