eˣrnesto @vedax

GLM-5.2 is now available on NVIDIA NIM with a free endpoint. The advantages: The model is fast and free. The downside: NVIDIA appears to be using a quantized version, the quality is lower than the officialAPI.

@AnthropicAI @MLStreetTalk 🤡

Anthropic hates Opensource AI because a model like GLM 5.2 being available for feee to download made their $1 Trillion valuation sound like a joke (which it is btw)

📢 The 5-Day AI Agents: Intensive Vibe Coding Course With @Google is now available as a Kaggle Learn Guide! Missed the live sessions from June 15 to 19? Access whitepapers, podcasts, codelabs, and recorded livestreams at your own pace.

To help accelerate neuroscience breakthroughs, we're releasing the full training code for Brain2Qwerty v1 and v2, and our partner, @bcbl_, is releasing the v1 dataset. Learn more and explore the artifacts here: go.meta.me/42ed9c

It's quite rational to regulate frontier API models, especially to get more transparency for the government, without regulating open-source AI. Here's why: 1. The most dangerous AI systems right now aren't open models. They're the large frontier LLM APIs distributed through coding tools and assistants, because: - They're built in secret behind closed doors and stay total black boxes. Zero transparency on what they can or can't do, with "safeguards" that blur everyone's ability to even analyze them. - They're built and controlled by a few profit-maximizing megacorps, concentrating unprecedented power in very few hands, with every incentive to downplay risks and overstate their safeguards. - They're distributed to hundreds of millions, maybe billions, of people and trivially easy to run. 2. Open-weight models are orders of magnitude less risky: - They're not as massively distributed or as easy to use, especially the big ones, as APIs and assistants. - We (including governments) can quickly and accurately analyze what they're capable of, and for now everyone confirms they're not as good as the APIs at doing bad things. - They're distributed to everyone, so defenders and law enforcement get as much access as attackers. The cost-benefit analysis of regulation is completely different too. Regulating frontier APIs is relatively easy and low risk while regulating open source would be much more complex, less efficient, and orders of magnitude more costly. Regulating frontier APIs would only potentially hurt a few megacorps, if it even hurts them, given all the marketing that it is already generating for them. They can afford armies of lawyers and absorb losing a few billion dollars, especially given they're on track to become some of the most valuable companies in history. Regulating open source, by contrast, would hurt the very people regulation is supposed to protect: small businesses, startups, researchers, nonprofits, universities, independent developers, and the broader public, while risking killing competition, slowing AI progress, and reducing transparency even more!

@KevinR4v @StarPlatinum_ what bear market?

Great to see the @NVIDIAAI Nemotron team put out a step-by-step guide for self-hosting Nemotron-3-Ultra 550B without a datacenter. Four compact DGX Spark boxes pool into a single OpenAI-compatible endpoint, served from vLLM's official out-of-the-box container. If you're building a private cluster of your own, don't miss this guide: 🔗 github.com/NVIDIA-NeMo/Ne… 🤖 Then wire that endpoint into your agent workflows, all on hardware you own.

I put together a new article on setting up local coding agents with open-weight models. Everything runs 100% locally. I thought it might be useful putting this together because many people asked me about my setup in the past, and I thought it would also motivate people to get started tinkering with local models for serious work (yes, things got incredibly capable this year with better LLMs and better harnesses). So, here's a walkthrough of how to connect a local LLM to a local coding harness (could be Claude Code or Codex, which you may already be familiar with). I also included some assessment notes that are useful as a checklist to select between and consider certain LLMs over others: - Checking RAM usage at long contexts to see if the model is suitable for real work - Measuring prefill and decoding tok/sec to see whether it's fast enough to not be annoying - Making sure the model has sufficient tool-calling capabilities in theory - Assessing whether the model can solve some more challenging tasks when used in a coding harness. Of course, there are always more specialized tools that can squeeze a bit more performance out of things, but I hope this is a good starter kit that stays flexible; that is you can easily switch to newer models as they are released or even tap into cloud models in your familiar harness if the current ones are not sufficient enough for a given task.

Wanna replace Anthropic/OpenAI? START WITH THIS The bible for running LLMs locally is now available online to read for free Covers what to use on - Laptop / edge / odd hardware - Mac-first workflows - Single RTX GPUs - 2-4+ NVIDIA / CUDA GPUs - General production serving - Long-context / MoE / routing - NVIDIA max performance - Cluster orchestration Software - llama.cpp - MLX / MLX-LM - ExLlamaV2 - ExLlamaV3 - vLLM - SGLang - TensorRT-LLM - NVIDIA Dynamo You should read this, and if you cannot now then you most definitely wanna bookmark it for later Opensource & Local AI FTW

Ahmad @TheAhmadOsman

2 months ago

x.com/i/article/2057…

35 196 2K 905K 4K

How did GLM-5.2 (Max) get to the top of Code Arena: Frontend? Looking at matched head-to-head on real-world web dev frontend tasks, @Zai_org's latest model takes a higher win share than its opponent in every pairing but one. - Beats every Claude Opus variant head-to-head: including Opus 4.8 (Thinking) and Opus 4.7 (Thinking) - Widest margins come against strong models: 61.0% vs Kimi-K2.6, 59.4% vs Sonnet 4.6 and 55.0% vs Opus 4.7 (Thinking). - The closest fight isn't from the very top: GPT-5.5 (xHigh), ranked #16, pushes it hardest at 41.7% to 40.0%. Opus 4.6 is the other (47.0% to 42.4%). There is a generational nuance here. Versus its own predecessor GLM-5.1, the latest model is tied: 45.5% - 45.5%. The jump shows up against the field, the one pairing without a clear win is in a direct rematch with 5.1.

Arena.ai @arena

3 weeks ago

Exciting news: GLM-5.2 (Max) ranks #2 in Code Arena: Frontend, with +29pt over Claude Opus 4.7 (Thinking) and only behind Fable 5! GLM-5.2 is the best open model vs Kimi-K2.6 and Minimax-M3 by a large margin. - #2 React and #4 HTML sub-leaderboards - Ranks as the top model in

187 544 5K 1.5M 1K

In our article, we explicitly laid out the criteria that would falsify the Bitcoin power law, either in the past or in the future. But falsification in the future would not mean that Bitcoin was never a power-law system. That is an important distinction. At this point, given that the work has passed peer review, it is a scientific result that Bitcoin has displayed scale-invariant properties for 17 years. Scale invariance is the core property of a power law. The question is not whether Bitcoin has behaved this way historically. The question is whether it will continue to do so. If Bitcoin stops following the power law, there are scientifically valid ways to determine that. Those are the kinds of criteria we described in the article: sustained structural deviations, breakdown of the long-term exponent, persistent failure to recover toward the scaling trend, and other similar tests. The “floor” everybody discusses is actually an anomaly. It sits roughly one sigma below the trend line, which is surprisingly tight given Bitcoin’s normal historical variability. So we are very far from saying that the power law is broken. That claim only makes sense if you understand the actual mathematics behind the model, not if you are just looking at a single support line and treating it as the whole theory. P. S. Notice the "floor" mentioned in the article is a 3 sigma floor (that is a order of magnitude away from the trend line as it should be for a scaling phenomenon).

GLM-5.2 from @Zai_org on ARC-AGI (Verified) - ARC-AGI-2: 22.8%, $0.25 - ARC-AGI-1: 77.0%, $0.19 Performance is comparable with GPT-5.4 & 5.5 (Low Reasoning Effort)

This is the strongest ARC-AGI-2 performance to date by an open-source model.

ARC Prize @arcprize

a week ago

GLM-5.2 from @Zai_org on ARC-AGI (Verified) - ARC-AGI-2: 22.8%, $0.25 - ARC-AGI-1: 77.0%, $0.19 Performance is comparable with GPT-5.4 & 5.5 (Low Reasoning Effort)

44 74 1K 344K 186

Kog open-sourced on @huggingface the 2B model that they used to show a model running at 3,000+ tokens per second. Very cool work! huggingface.co/blog/kogai/kog…

My family and I are exposed to all sorts of threats, restrictions and hateful rethoric. Yet we would say and do everything we have said and done against the genocide, its ideologues and profiteers, again and again. Forever proud to stand on the side of Justice.

susan abulhawa | سوزان ابو الهوى @susanabulhawa

2 weeks ago

Hillel Neuer, who denies zionist influence (control) in the US, brags about destroying Francesca Albanese's ability to conduct normal daily life transactions, and how they then went after her husband too.

175 4K 9K 561K 643

Excited to partner with @OpenRouter ⚡ Products like OpenRouter Fusion and Sakana Fugu have sparked a serious conversation about dependency and resilience in AI. I believe this is just the start of a great architectural shift to come in AI development.

Sakana AI @SakanaAILabs

a week ago

Fugu-Ultra is now live on @OpenRouter! ⚡ We share a core vision with the OpenRouter team: the future of AI isn’t a single monolithic model, but the collective intelligence of the world’s best models working together. Try it: openrouter.ai/sakana/fugu-ul… 🐡

61 187 1K 223K 314

The best way to understand a complex system is via edge cases and failure modes, because they define the contour of the system.

📣📣 Meet Qwen-AgentWorld — a native language world model that simulates 7 agent environments (MCP, Search, Terminal, SWE, Web, OS, Android) within a single model. Environment modeling is the training objective from day one, not a post-hoc adaptation. 🤔 LLMs are trained to be better agents — better at acting in environments. But nobody has trained them to model the environments themselves. 🗺️ Our roadmap: investigate how language world modeling can push the boundaries of general agent capabilities, along two routes: 1️⃣ Build a foundation model for environment simulation — outperforming Claude Opus 4.8 and GPT-5.4 on AgentWorldBench 2️⃣ Investigate how world modeling enhances agent training: 🔬 Controllable Sim RL (agentic RL with LWM as environments) surpasses training in real environments 🧠 Learning to predict environments (LWM warm-up) makes agents stronger — remarkably, even without any agent-specific training, this predictive knowledge transfers to agentic tasks with zero fine-tuning 📑 Paper: arxiv.org/abs/2606.24597 📖 Blog: qwen.ai/blog?id=qwen-a… 💻 GitHub: github.com/QwenLM/Qwen-Ag… 🤗 HuggingFace: huggingface.co/collections/Qw… 🧩 ModelScope: modelscope.cn/collections/Qw…

1-bit GLM-5.2 GGUF vs. Claude 4.8 Opus vs. GPT-5.5 We gave 3 models the same prompt and compared one-shot outputs. The 1-bit GLM-5.2 GGUF ran locally on a Mac Studio M3 Ultra with 256GB RAM at ~21.6 tok/s. Which output do you like best? GGUF: huggingface.co/unsloth/GLM-5.…

Unsloth AI @UnslothAI

2 weeks ago

GLM-5.2 can now be run locally!🔥 The 2-bit model retains ~82% accuracy after we shrunk it from 1.51TB to 238GB (-84% size). Run on a 256GB Mac or RAM/VRAM setups. GLM-5.2 is the strongest open model to date. Guide: unsloth.ai/docs/models/gl… GGUF: huggingface.co/unsloth/GLM-5.…

272 893 7K 1.9M 4K