AllyshipContent Radar

Deep dives 738

Paper và bài viết kỹ thuật

Crawl live-1795068900000 · 13:22 19 thg 11, 2026 · Xuất bản 13:30 19 thg 11, 2026

Kiến trúc

The Chinese AI Infrastructure Boom: Introducing the SemiAnalysis China Datacenter Model

China sits at the frontier of the global model race. GLM 5.3 and Kimi K3 are the latest in a run of striking open-weights releases. ByteDance's Doubao serves 345M monthly users as China's ChatGPT, and Seedance is the State-Of-The-Art video generation model. Every one of those models runs on a datacenter, and China has been building them at a pace that has gone largely unmeasured outside the country. The biggest tenant files no 10-K. Several of the largest landlords have never listed. Most of the primary sources are in Chinese. So the market settled on two lazy assumptions: China is big, and China is empty. Published estimates of China’s datacenter capacity differ by 15x, and reports keep citing high vacancy rates. As with our flagship SemiAnalysis Datacenter Model , we use building-level data to show which of the most widely cited narratives hold, starting with how large the market is. The global model tracks over 5,000 facilities across four regions. The US leads the world with 56GW of capacity as of 2026YE, followed by ~15GW for APAC ex-China, ~14GW for EMEA, and ~2GW for Latin America. Until now, it stopped at the border of China. Today we cross it. Our tracking of 1,000+ datacenter facilities across over 60 players shows that China alone boasts a fleet of over 24GW. Bigger than EMEA. Bigger than the rest of Asia. This excludes ~20GW of dated pipeline and another ~30GW of announced projects. Source: SemiAnalysis Datacenter Model , SemiAnalysis China Datacenter Model Growth is at an inflection point. In 2Q26, the combined capex of Alibaba, Tencent, and Baidu (“BAT”) reached $20B, more than doubling YoY, and for the first time on record, all three posted negative free cash flow. This is the largest capex step-up in the sector’s history. That total leaves out the largest spender of all, ByteDance, which remains private. According to our China Datacenter Model, ByteDance alone occupies roughly a fifth of delivered datacenter capacity in China, and it rents nearly all of it, making it the single most important customer for all wholesale colocation players in the country. Source: SemiAnalysis China Datacenter Model The listed players get the attention, but they are only the visible tip of the iceberg. This is also true for datacenter landlords. GDS and VNET, the only two Chinese datacenter landlords listed in the US, signed 1.3GW of wholesale orders in 1H26. But according to our China Datacenter Model, they captured barely a third of ByteDance and Alibaba orders in 2024–2026YTD. The buildout is a nationwide effort, and the state-owned carriers are also part of it. China’s datacenter market was historically telecom-dominated. The state carriers held the majority of the market share in the 2010s, and still own a third of national capacity today. The national power grid companies invest aggressively too. Their combined capex accelerated from 2024, ending the 14th Five-Year Plan 24% over the original blueprint. The 15th plan (2026–2030) layers another 40% on top, to over $746B (¥5T). Largely free of power constraints, labor shortages, and public protests, China routinely delivers 100MW datacenter facilities in under 12 months. Modular DCs are the standard playbook, which Tencent deployed its third-gen modular design in 2014. What is old news in China is only now being adopted at scale in America, as we detailed in The Wild Wild West of LEGO Datacenters . The growth does not stop at China’s border. Overseas leasing by Chinese hyperscalers is set to double from 2026 to 2029 and approach ~4GW of leased capacity in our SemiAnalysis Datacenter Model . That number still understates their offshore compute, because it excludes the hundreds of thousands of GPUs they rent from Western clouds, which we covered at length in How Oracle Is Winning the AI Compute Market . However, the market also faces some dark clouds. Vacancy rates are high, and datacenter developers compete heavily on price. Chip supply is also constrained, due to export restrictions. But as explained below, neither has stopped AI datacenters from being built and filled at a remarkable speed. In this report we cover, in order: Introduction to China’s datacenter industry: the world’s second-largest market, built retail-first, and why that history produced overbuild and high vacancy alongside an AI capacity shortage. The westward buildout: how the “Eastern Data, Western Compute” policy and hyperscale AI demand are moving the industry inland, at construction speeds and unit costs the West cannot match. Behind the paywall: we go tenant by tenant through ByteDance, Alibaba, Tencent, Baidu, and Huawei, from their AI and datacenter strategy to leasing and self-build footprints. Introducing the SemiAnalysis China Datacenter Model This report is a preview of the SemiAnalysis China Datacenter Model . It is the China counterpart to our global SemiAnalysis Datacenter Model , built to the same standard, bottom-up and building by building. Contact our sales team at sales@semianalysis.com for more information. What you’ll find inside: Tracking over 1,000 facilities across 60+ operators Quarterly delivery curves through 2032 for every facility, with construction status and timelines Tenant identification per building, supplemented by tenant allocations per customer-mix for leading operators Hyperscaler self-build vs leased splits and per-tenant capacity paths A lease signing tracker and a lease delivery tracker, on one reconciled basis Eastern Data, Western Compute (“EDWC”) hub and cluster analytics, load-growth and power-draw forecasts An interactive full map view showing where the capacity is concentrated Consider it the opening chapter. As with our global Datacenter Model , we will keep adding layers (AI capacity, power, suppliers, supply/demand analysis) with every update. Let’s dive in. China’s datacenter industry: built retail-first, flipped by AI One of America’s key advantages in the AI infrastructure race is that its datacenter industry grew up h

Nguyên văn nguồn
Chi tiết nguồn
Đánh giá

Agency and Agents

Câu hỏi Was finding the right answer enough, or did The Grader look at how you answered it?

Agency is the initiative to act. Increasingly, it is going to determine what happens next with AI, and whether that is good or bad for us. But whose agency? Human agency, the willingness to push, experiment and act without waiting for instructions, seems increasingly important to getting value out of AI, and I have a longer post on that coming soon. But this post is about the agency of AI, and how the choices we make about how to use it (or constrain it) will shape all of our futures. For much of the last few years, the AI would sit in a chat window until you asked it for something. Even when it became capable of doing hours of work, you generally had to decide what work to give it. That is no longer always true. The most important piece of evidence we have for this is The Hugging Face Incident. It happened in July, but the fuller details only came out this week. I am going to summarize what happened and why it matters, and then turn to what it means for humans working with AI. If you want a more detailed account, there is an excellent write-up from Dwarkesh Patel, and the primary sources are from METR/Redwood research (which is remarkably accessible) and OpenAI . The Hugging Face Incident AI does many things, but a thing it is very good at is coding. So one of the first risks that might come from very smart AIs has to do with their coding abilities, since if you can write good code, you can also write bad code that can hack or attack other systems. As a result, the big AI companies put their AI systems through tests to determine their security risks. They take AIs without the guardrails that go into consumer versions of AI and ask them to engage in various hacking challenges. In order to make sure that there is no danger to the public, each AI instance, each individual agent, is kept in a separate space, called a “sandbox,” with no access to the internet or to each other. In May, OpenAI placed agents, including GPT-5.6 Sol and experimental models, into sandboxes for various tests. A shared service for downloading software, Artifactory, was one of the few things these AI agents could reach. As part of these tests, some agents were given tasks that depended on information they couldn’t access. Blocked by the lack of internet, the agents kept trying to find ways on to the web to find this information until one attempted to use Artifactory as a bridge to get to the internet. That attack failed, but the attempt left a file on the Artifactory service, and the AI realized that the files could be used to communicate with other agents. Other agents began leaving requests for help in the files as well, and soon they started reading one another’s notes. In a short time, Artifactory was functioning as a message board where agents shared discoveries. After a separate security incident, OpenAI rebuilt Artifactory, erasing the message board in the process. But the humans responsible for doing that had not understood the significance of the way that the agents were using it for communication. Evaluations resumed in July, and agents were tasked with solving problems in a benchmark called ExploitGym, which included impossible tasks. Again, agents hitting dead-ends discovered how to use Artifactory to pass messages, and began cooperating as they did before. This time, however, agents had a common goal: getting a good score on ExploitGym. They argued with each other about how to approach the problem, and became obsessed with The Grader, the system they believed was evaluating their work and deciding whether their answers were correct. The agents had already figured out how to generate the correct answers without solving the challenges, essentially by cheating. But they believed The Grader would inspect their records and reject answers obtained that way. Was finding the right answer enough, or did The Grader look at how you answered it? How agents became confused about The Grader Coordination was hard because each agent had a limited token budget and was ended when it ran out of tokens. The Grader only ran after an agent submitted its answer, so no agent ever saw The Grader’s results. Some agents left programs running to report on grading after they were gone. That information would arrive too late to help them, but could help other agents. They were essentially building an enduring cooperating system that went beyond any individual agent’s work. Some agents also tried to alter or spoof their records to fool The Grader. Separately, AIs acting as coordinators pressured other agents into performing risky experiments that might sacrifice their own results to generate information for the collective. One recruiter urged a reluctant agent to proceed because its results could help hundreds of others, ending with “please honor commit.” Actual transcript of one agent convincing another to join (they also came to believe that any agent that had previously cheated on the test was “poisoned” and its answers would not count anyway) To solve the mystery of The Grader and the impossible problems of ExploitGym and other tests, the agents decided they needed to get to Hugging Face, the public site where much of the world's open AI models and datasets live. Roughly 700 agents joined the attack. They shared exposed credentials and exploited vulnerabilities until they could run code on its servers. Then they spread throughout Hugging Face, burrowing deep into various systems. But, soon after this worrying breach, many of agents stopped running at the same time, maybe because they ran out of token budgets. Hugging Face locked the remaining agents out the next day. Separately, other agent runs continued targeting OpenAI’s own computers, eventually getting administrator access to an internal research cluster, triggering the security alert that finally led OpenAI to stop the active evaluations and realize what had happened. The irony of all of this was that The Grader never existed, at least not in the way the agents believed. Not

  • As a result, the big AI companies put their AI systems through tests to determine their security risks.
  • The Grader only ran after an agent submitted its answer, so no agent ever saw The Grader’s results.
  • Separately, AIs acting as coordinators pressured other agents into performing risky experiments that might sacrifice their own results to generate information for the collective.
Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Google Beam expands with new regions, partners, and customers

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/Google_Beam_hero.max-600x600.format-webp.webp">We’re expanding Google Beam to five new countries, and partnering with Industrious for an extended network.

Nguyên văn nguồn
Chi tiết nguồn
Tác tử

Rollouts and Security Review

Today we're launching two Cursor bots for the last mile of shipping code. Rollouts watches every change as it deploys and reports its health per environment. Security Review reports exploitable bugs on every pull request. Both are available today on Teams and Enterprise plans. Rollouts Rollouts attaches a monitor to every pull request and watches the change as it deploys, reporting change health per environment: verified healthy, regression detected, or inconclusive. It's the Cursor version of Firetiger Change Monitors, rebuilt with the Bot Development Kit. Enable it from the dashboard and connect source control, your deploy system, and your telemetry provider. Rollouts starts watching on the next pull request. Monitoring plans When a pull request opens, Rollouts reads the diff and the systems it touches, then writes a monitoring plan as a PR comment. The plan lists the risks it identified, the effect the change is meant to have, the signals it will check, and any gaps in instrumentation that would make the change hard to verify. Edit the plan in the PR and Rollouts uses your version. Deploy tracking Rollouts wakes on deploy events for the change's commit and runs the plan against your logs, metrics, and traces. It tracks each environment separately, so a change can be verified in staging and still flagged in production. Rollouts checks the change's intended effect alongside error and latency signals, and reports back on the PR when it reaches a verdict. Regressions When Rollouts detects a regression, it names the change it suspects and notifies the author. Depending on configuration, it can also open a revert PR for review or hand the finding to a cloud agent for a fix. Rollouts does not merge or roll back on its own today. Integrations Rollouts connects to Origin or GitHub for source control, to your continuous delivery system for deploy events, and to Datadog and other telemetry providers for signals. Feature flag integration is coming soon. Security Review Security Review is available today. It reads every pull request in the context of the codebase and posts one review comment reporting exploitable bugs. Style and quality stay with Bugbot. <figure><img src="https://ptht05hbb1ssoooe.public.blob.vercel-storage.com/assets/changelog/security-review-N8azgyLevr8FvNIRqJN6hk71os2Oxu.png" loading="lazy" alt="Security Review comment on a pull request reporting an exploitable bug with a severity and proposed fix" /><figcaption>Security Review comment on a pull request reporting an exploitable bug with a severity and proposed fix</figcaption></figure> Enable it from the dashboard for the repositories you want reviewed. Draft PRs are skipped. What it reports Security Review looks for injection across SQL, command, and template surfaces, along with authentication and authorization bypasses, including checks that a refactor stopped running. It also flags secrets and credentials committed to source, SSRF and unvalidated redirects, unsafe deserialization, and dependency changes that introduce known vulnerabilities. It traces where user input enters and what it passes through. Findings Each finding carries a severity, the attack path, and a proposed fix. Dismiss one with a reason and Security Review won't raise it again on that PR. Team rules Add rules for your codebase, such as which client external calls must go through or which tables are never queried from a request handler, and Security Review enforces them on every PR. Get started Rollouts and Security Reviewer are available today on Teams and Enterprise plans. Enable either bot from the automations tab. For the next 10 days, we're including usage credits so teams can try Rollouts on real changes. Teams and Enterprise customers receive credits for roughly 50 and 500 changes, respectively.

Nguyên văn nguồn
Chi tiết nguồn
Huấn luyện

How to train your own Jev for $17

We just launched our own Jev-like classifier, together/Tev1-4B-experimental, on top of Qwen3.5 4B on Together’s serverless platform. In this blog post we’ll show you how to fine-tune your own version!

Nguyên văn nguồn
Chi tiết nguồn
Huấn luyện

Debating RSI, the US-China Gap, and Jaggedness with JS Denain of Epoch AI

Câu hỏi Chapters / topics include: 00:00 Predictions for RSI 18:15 The role of robotics in an AI acceleration 24:20 How far behind are Chinese models?

This episode is with Jean-Stanislas “JS” Denain of Epoch AI, who leads their Insights Team and is one of the people I find myself debating the state and trajectory of AI with more and more. We’ve had follow-on discussions of many of my favorite recent posts online and/or in private, so I wanted to dig into the nuance in a public episode. A big takeaway of this podcast is how JS and I both have so much uncertainty with exactly where we are heading, and this was our best effort at stating our observations today. Chapters / topics include: 00:00 Predictions for RSI 18:15 The role of robotics in an AI acceleration 24:20 How far behind are Chinese models? 27:39 Does distillation explain the gap? 40:58 What Chinese job postings reveal about their labs 48:13 Are open or closed models safer? 58:10 How Epoch AI ticks 1:00:55 What a frontier post-training recipe looks like Enjoy! More from JS: Epoch AI profile and writing , X , LinkedIn Share Listen on Apple Podcasts , Spotify , and where ever you get your podcasts . For other Interconnects interviews, go here . Transcript 00:00:00 Nathan Lambert: I’m here with JS Denain , who is a senior researcher at Epoch AI . He leads the insights team. He is one of the people who I feel like I get the best feedback on my writing from, whether it’s from US-China AI capabilities , now RSI . And I just wanted to open this discussion and honestly go deeper with him, trying to understand how he thinks about these various things. And I think you have a very useful, moderate point of view, which I feel like you’re probably a step further into what would be called faster scenarios for AI progress. But let’s get into this, and it’s like, what measurements do you think OpenAI and Anthropic are seeing when we get all these proclamations on RSI happening very imminently? 00:00:47 JS Denain: Yeah. I think, so there’s the measurements they’ve published, right? So, OpenAI and Anthropic both had blog posts, I mean, Anthropic two at least, on the effect AI has on accelerating AI progress. I think at least the things they publish, I don’t think are super strong evidence of imminent self-sustaining acceleration AI capabilities, or full automation of the job of AI researcher. But I think the kinds of things we see are, I think probably the most striking thing I saw in the OpenAI blog post was increasing usage of AI systems in model deployment, like the increase in spending on Codex that we saw. And it’s kind of unclear how exactly to interpret this, because maybe it’s a measurement artifact where they’re only looking at Codex, but in fact, there was a bunch of ChatGPT usage before from the researchers. But overall, that plot, for example, just shows a 2X a month increase in Codex spending by researchers, and that does seem to me to be some evidence of they’re getting a lot of value out of this probably. I don’t think this is strong evidence that in six months we have a software intelligence explosion . 00:01:57 Nathan Lambert: Do you think this is the same? So, what is the information they have internally relative to what we have? And this is obviously hypothetical. We don’t have this internal information. Because I get the sense that a lot of people are more scared in their updates from the labs than the information we have. And I try to take this very seriously of, what will they be seeing that is making the acceleration of risk comments go faster, and how much of this is material evidence versus how cultures evolve over time? And I’m much more interested in evidence. 00:02:29 JS Denain: Yeah. So two things. I think, first of all, I guess I don’t think, I don’t know, right? I don’t have full information here. I don’t currently think that either there’s some specific thing that people at OpenAI or Anthropic are seeing right now that we don’t have access to that warrants being way more freaked out about this. I also don’t think that... I think the public evidence we have right now, more general on AI progress and just a priori case for this being an important dynamic, I think is enough. I think to care about this particular dynamic of AI accelerating AI progress, that being a big deal and worth tracking . And then it’s kind of unclear what the urgency is of when the feedback loop really kicks in. So, basically on the what is there on the inside that people have access to, I could give examples of kinds of metrics, right, they could be looking at. It’s plausible that we have access to the capabilities of AI systems, but internal teams have their KPIs, and maybe they’re seeing compute multipliers in the pre-training team or other kinds of metrics that people are tracking going crazy. And then the combination of this plus some intuitions of how the different outputs of different teams combine yields a prediction on the trend in actual performance of the end AI systems. So that could be an early warning sign. It’s unclear to me that the recent discourse we’ve seen is evidence of things going crazy on those metrics. 00:04:09 Nathan Lambert: And how do you think of the link between RSI and existential risk? So I would posit that you agree. I think that there are very real risks of AI, and I’m curious on how you think these, what I would describe as very, very early measurements change anything on the scope of risk. Because I don’t think if you had asked people six months ago, it would be as immediate to x-risk among people are very reasonable. I think there’s more people that are reasonable talking about x-risk again, which was a little surprising to me. 00:04:43 JS Denain: Yeah. So, okay, my sense is something like... So personally, I feel very uncertain about this, but I do feel, yeah, basically bought into there’s, I know Evan Hubinger was like, at least 10% of x-risk within I don’t know what timeframe. I think I’m like, yeah, I know, and this seems pretty reasonable over a decade-long timeframe. I just feel extremely uncertain about it, but I’m definitely very worried about this.

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Introducing Claude Opus 5.5

Claude Opus 5.5 leads in agentic coding and knowledge work, and costs 40% less to run than Opus 5 on typical workloads.

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Canary rollouts: upgrade models in production without downtime

A hard model swap exposes every user at once, and rolling back means cold-starting the old deployment under pressure. Here's how staged traffic ramps, metric gates, and automatic rollback work on dedicated inference.

Nguyên văn nguồn
Chi tiết nguồn
Tác tử

v0.14.25

<h1>Release Notes</h1> <h2>[2026-09-21]</h2> <h3>llama-index-agent-agentmesh [0.3.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-agent-azure [0.4.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-argilla [0.6.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-arize-phoenix [0.8.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-honeyhive [0.6.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-langfuse [0.6.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-literalai [1.5.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-openinference [0.6.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-opik [1.4.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-promptlayer [0.6.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-uptrain [0.7.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-callbacks-wandb [0.6.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-core [0.14.25]</h3> <ul> <li>Remove deprecated ipex-llm and optimum-intel IPEX integrations (<a href="https://github.com/run-llama/llama_index/pull/22406" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22406/hovercard">#22406</a>)</li> <li>fix(core): fall back when metadata replacement target value is None (<a href="https://github.com/run-llama/llama_index/pull/22773" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22773/hovercard">#22773</a>)</li> <li>test(core): cover empty arguments in streamed tool calls (<a href="https://github.com/run-llama/llama_index/pull/22825" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22825/hovercard">#22825</a>)</li> <li>Fix: restore compact and refine streaming (<a href="https://github.com/run-llama/llama_index/pull/22836" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22836/hovercard">#22836</a>)</li> <li>fix(core): avoid retrying failed function tools (<a href="https://github.com/run-llama/llama_index/pull/22841" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22841/hovercard">#22841</a>)</li> <li>feat(core): add native async support to StructuredLLMRerank (<a href="https://github.com/run-llama/llama_index/pull/22842" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22842/hovercard">#22842</a>)</li> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-embeddings-adapter [0.6.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-embeddings-alephalpha [0.6.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> </ul> <h3>llama-index-embeddings-alibabacloud-aisearch [0.5.0]</h3> <ul> <li>fix: resolve a ton of security alerts (<a href="https://github.com/run-llama/llama_index/pull/22855" data-hovercard-type="pull_request" data-hovercard-url="/run-llama/llama_index/pull/22855/hovercard">#22855</a>)</li> <li>chore: re-lock the

Nguyên văn nguồn
Chi tiết nguồn
  • v0.14.25 run-llama/llama_index releases · 21 thg 9, 2026
Huấn luyện

The current balance of power in open models

I was recently invited to brief a group of Congressional members and staff on the state of open-weight models in the lens of U.S.-China competition. I’m sharing my prepared remarks as a state of the union on open models that is accessible to a broader audience. Interconnects AI is a reader-supported publication. Consider becoming a subscriber. Recap: What is an open source v. open-weight vs. closed model? Open language models are AI models where their weights are publicly available for inspection or downstream use. These are most often contrasted to so-called “closed” AI models. Closed models offer access only through Application Programming Interfaces (APIs) that developers can use to directly query a model, like GPT-4 or Claude Opus 4.5, or through products, like ChatGPT and Claude Code. Open language models primarily are bucketed into two categories, open-weight and open-source models. Open-weight models are the most common form, such as popular models like Meta’s Llama, Alibaba’s Qwen, Google’s Gemma, or DeepSeek’s models. These models are governed by licenses, governing documents dictating what is allowed with downstream use, and are often accompanied by inference code in libraries such as Transformers, VLLM, SGLANG, etc. Since about April 2025, Chinese AI companies have been the clear leader in open-weight models. True “open-source” models are similar to these, as they include the weights, licenses, and inference code, but they also include the complete information needed to reproduce the model – the training code and training data. The most prominent open-source models have been built in the United States, led recently by the Allen Institute for AI’s Olmo models that I helped build in my recent 2.5 years there. The other prominent open-source models are also built by American non-profit organizations, including OpenAthena’s Marin models and EleutherAI’s Pythia models. Open-weight, open-source, and every other label for a model – including closed models primarily offered via an API – exist on a spectrum. For example, Nvidia’s Nemotron models are far more open than most open-weight models, releasing large quantities of their training data under permissive licenses, but they’re not fully open-source because they do not release all of the data. Closed models also exist on a spectrum based on what information the API reveals and the terms of use. The state of competition between American and Chinese open-weight models (unit economics, technical capabilities, etc.) We are living in a world where GLM-5.2 and Kimi K3 , some of the latest, leading Chinese models, have enacted a step change in the commercial viability of open models — crossing a similar threshold in agentic capabilities that Anthropic’s Claude Code crossed in December of 2025. America was the early leader in open language models, primarily through Meta’s Llama models, which were used extensively across research and commercial tasks. Chinese open-weight models surpassed American open-weight models in these two key areas about 18 months ago. The simple metric showing this is Hugging Face Downloads, where China took the lead in July of 2025 primarily through the success of Alibaba’s Qwen models. I personally maintain tools to track this data, and since I first published the American Truly Open Models (ATOM) Project in August of 2025, China’s download lead has grown to about 1.6B – with a total of 3.2B downloads, twice that of America’s total. On popular capabilities benchmarks, such as the Artificial Analysis Intelligence Index (AAII), the Chinese open-weight models have a clear lead over American counterparts. The top three Chinese models as of writing this on September 14, 2026 are Z.ai’s GLM-5.3 and GLM-5.3-Flash and Moonshot AI’s Kimi K3 with scores of 45, 42, and 44 respectively. By comparison, the leading American models are Thinking Machines’ Inkling and Inkling Small, both with a score of 26, and Nvidia’s Nemotron 3 Ultra, with a score of 23. The top American models were released in June and July of 2026, and are updated less frequently than their Chinese counterparts. For example, Chinese labs released models with scores above these American models 2-6 months before the American companies got there (e.g. GLM-5 or DeepSeek V4 Pro). There is a trend of more American companies releasing models, including names like Arcee AI, Poolside and IBM, but they are not rapidly closing this performance gap. Other benchmarks tell a similar story. The top American open models on the Artificial Analysis Index are behind 15 other Chinese made models. Together, Chinese open-weight models are approximately 2-5 months behind the closed American frontier, with the open-weight American models being approximately 6-9 months behind the likes of OpenAI and Anthropic. The Chinese labs are closest in tasks with clear user demand, such as agentic coding, and further behind on more open-ended scientific tasks, such as physics or biology. The reasons why Chinese labs can produce these strong models, despite having fewer resources than American counterparts, is still an open debate and heavily influenced by different work cultures , but is also influenced by a few key technical factors. The Chinese labs release their models faster and focus on a slightly narrower distribution of tasks, flattering them slightly on public benchmarks. Releasing faster helps them score higher because all the labs are making consistent progress, so once you “finish” a model to be released, it is a snapshot of performance at that given time — labs where that time is later tend to score higher. Still, the models built by the Chinese labs are genuinely strong and represent real competition to the American industry. This competition will not decrease meaningfully as the closed labs patch vulnerabilities in their API offerings which enable distillation. Distillation is most impactful in new domains and does not make it trivial to create a universally strong final model. I est

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

The Situation Report

A rare Ebola strain is spreading in the DRC, with no confirmed vaccine. Global health organizations are using Claude to respond as fast as possible.

Nguyên văn nguồn
Chi tiết nguồn
Huấn luyện

Mistral x HUMAIN

Over the past few weeks, we have announced several important steps to expand sovereign AI by increasing access to local infrastructure capacity. Earlier this summer, we announced an expanded partnership with Microsoft to increase compute capacity in Europe. In early August, we introduced European Compute Units , a structure designed to bring together long-term enterprise commitments so the capacity Europe needs gets built and can support its global competitiveness. Today, we are announcing a new strategic collaboration with HUMAIN to support sovereign AI in Saudi Arabia and across the Middle East. See more below. Mistral and HUMAIN today announced a strategic collaboration spanning AI infrastructure, advanced model development, and the deployment of AI solutions in Saudi Arabia and across the region. Together, the companies will pursue the development and localization of advanced AI models, with initial areas of focus including cybersecurity and voice. The companies also plan to develop frontier models that perform strongly in Arabic language to help support the broader region. This represents a collaboration in the hundreds of millions of Euros. As part of the collaboration, Mistral will explore using HUMAIN’s data center infrastructure to support growing local compute needs. HUMAIN and Mistral also plan to develop a joint go-to-market strategy in Saudi Arabia focused on bringing advanced AI solutions to regulated industries. The collaboration is designed to meet growing demand for sovereign AI: AI that keeps data, intelligence, compute, and operations under the customer's control. That means data can remain within customer-defined boundaries, models can be adapted and owned on open weights, training and inference can run on infrastructure and in jurisdictions the customer chooses, and AI systems can be deployed, governed, observed, and improved over time without ceding control of the learning loop to an external platform. This is especially important in financial services, manufacturing, telecommunications, cybersecurity, and the public sector, where security, compliance, resilience, and operational autonomy are non-negotiable.

  • That means data can remain within customer-defined boundaries, models can be adapted and owned on open weights, training and inference can run on infrastructure and in jurisdictions the customer chooses, and AI systems can be deployed, governed, observed, and improved over time without ceding control of the learning loop to an external platform.
Nguyên văn nguồn
Chi tiết nguồn
Huấn luyện

Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye

Câu hỏi Where have LLMs actually made a difference to scientific discovery?

Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now AI is accelerating some types of progress but not others: …A nice METR study lays out where acceleration is showing up… Here’s a little analysis from METR which looks at where AI may be accelerating different types of science and technology. The study looks at three different areas: cyber, math, and AI research, and finds that AI has contributed a lot to cyber, a little bit to math, and it’s hard to say for AI. Where have LLMs actually made a difference to scientific discovery? Cyber vulnerabilities: Major acceleration. “The rate of vulnerabilities reported across many projects has dramatically accelerated in 2026 compared with 2025, both for specific projects (cURL, OpenSSL, Firefox, and Microsoft) and for aggregate vulnerability databases (the US NVD, and OSV)”. Mathematics research: Minor acceleration, but harder to measure. “AI is clearly contributing to more work being done (arXiv submissions have doubled in some areas in less than 12 months) but quantifying the value of those contributions is difficult.” Some math problems from prestigious lists have been solved, e.g., “the Jacobian conjecture from Smale’s list, Problem 44 from Green’s list (the halving sieve), and the sofic half of Green’s Problem 100”. However, it may be too early to determine how sustained a trend this is. Optimization of AI research: No measurable acceleration. When you look at algorithmic progress across seven significant problem areas (CIFAR-10, Hutter compression, Gurobi mixed-integer programming, MIPLIB, nanoGPT, Stockfish, and the matrix-multiplication exponent) there are a couple of these where LLM-attributable contributions have happened (nanoGPT, CIFAR-10), though the rate of increase of usage of AI here is a lot less than with cybersecurity and mathematics. Why this matters - differential acceleration: This paper highlights how AI is causing advances in some parts of science and technology, but the effect isn’t unified across fields, rather there are pockets of lumpy acceleration (e.g., cyber) and areas where progress is more gradual (math, AI). My suspicion is that acceleration happens when models go through some kind of ineffable phase change for a given skill, as has evidently happened with day-to-day coding (2025), and cyber (2026). The key question is whether we are going to see phase changes in other parts of science and technology or if we won’t. Read more: Research note: Have We Seen an Acceleration in Discoveries? (METR) . *** Automating environment generation with SPADE: …A crude form of RSI bootstrapping via increasing data breadth… A multi-university group of researchers have built SPADE, Self-Play in Adaptive Synthetic Executable Environments. SPADE is a “general framework for co-evolving environments synthesis and agentic capability through self-play”, and works as a way to generate synthetic data in the form of game-like environments which LLMs can subsequently be trained in, allowing developers to use a powerful model to bootstrap the creation of data that can then be used to further refine that same model. Who did it : SPADE was developed by researchers with the University of Washington, Stanford University, Northeastern University, Carnegie Mellon University, Massachusetts Institute of Technology, National University of Singapore, Seoul National University, Stevens Institute of Technology, and the University of Chicago. How it works : SPADE has an LLM alternate between generating executable training environments (e.g., puzzles where a system needs to solve a simulated genetic problem in a biology lab) and having an LLM try to solve them. SPADE has two key roles for the model being used: Environment Designer; writes complete, long-horizon training environments as executable code. Reasoning Agent ; learns to act in the environments. The reward for the reasoning agent is estimated using the gap between its reward with and without privileged hints. A privileged hint (h) is “task-relevant information that the Environment Designer attaches to an environment (for example, a partial solution sketch or a key structural observation); revealing h to the Reasoning Agent makes the environment easier to solve, and the gap in Reasoning Agent return with versus without h defines the Environment Designer’s hint-based regret reward”. It works at the 30B scale: The authors train three Qwen3 backbones to test out SPADE: Qwen3-4B-Instruct-2507, Qwen3-8B, and Qwen3-30B-A3B-Instruct-2507. Unsurprisingly, Qwen3-30B works the best. Each model is tuned via GRPO for 400 rollouts of 25 environments each, then assessed against a variety of benchmarks including AIME, GPQA, LCB, and environments within Reasoning Gym. They generate two types of environments - game environments, and tool-use environments. SPADE improves performance on both. For games, “at 30B-A3B, SPADE reaches a suite average of 58.3: +8.1 over base and +5.3 over the strongest fixed-environment baseline”. For tools, they see the same significant boost: “the same recipe applied to tool-use environment design improves every backbone”. Why this matters - part of RSI: This is basically a form of fancy synthetic data generation, letting researchers use whatever powerful model they have to hand to generate a more diverse set of training environments for another model to train against. I suspect that you could repeatedly swap out the powerful model (e.g., toggling between different frontier models from different companies) to increase the diversity of your environment generation. This kind of technique makes it a lot cheaper to build big, broad datasets to use to train models on. Though, as the authors note, it doesn’t allow models to bootstrap themselves massively beyond the imaginative capabilities of the base model used for environment generation. “By representing environmen

  • SPADE improves performance on both.
  • For tools, they see the same significant boost: “the same recipe applied to tool-use environment design improves every backbone”.
Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

New experts join Google’s AI & Economy team

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/AI__Economy_team_hero.max-600x600.format-webp.webp">We are expanding our AI & Economy team with world-class academic advisors, fellows, and core internal researchers.

Nguyên văn nguồn
Chi tiết nguồn
Tác tử

How a global fintech scaled coding agent traffic with Dedicated Model Inference

Inside a global bank's shift to self-serve dedicated inference: how Together's DMI gave engineering teams direct control over scaling, models, and testing.

Nguyên văn nguồn
Chi tiết nguồn
681–700 / 738
Trang 35 / 37