Wei Ping

Distinguished Scientist & Director @NVIDIA. Work(ed) on post-training, reasoning, agentic coding, and multimodality.

[Wei Ping @_weiping]

🚀 Introducing Nemotron-Cascade 2 🚀

Just 3 months after Nemotron-Cascade 1, we’re releasing Nemotron-Cascade 2: an open 30B MoE with 3B active parameters, delivering best-in-class reasoning and strong agentic capabilities.

🥇 Gold Medal-level performance on IMO 2025, IOI 2025, and ICPC World Finals 2025:
• Capabilities once thought achievable only by frontier proprietary models (e.g. Gemini Deep Think) or frontier-scale open models (i.e. DeepSeek-V3.2-Speciale-671B-A37B).
• Remarkably high intelligence density with 20× fewer parameters.

🏆 Best-in-class across math, code reasoning, alignment, and instruction following:
• Outperforms the latest Qwen3.5-35B-A3B (2026-02-24) and even larger Qwen3.5-122B-A10B (2026-03-11).

🧠 Powered by Cascade RL + multi-domain on-policy distillation:
• Significantly expand Cascade RL across a much broader range of reasoning and agentic domains than Nemotron-Cascade 1, while distilling from the strongest intermediate teacher models throughout training to recover regressions and sustain gains.

🤗 Model + SFT + RL data:
👉 https://t.co/4QJqfTOt6I

📄 Technical report:
👉 https://t.co/dFC00m6RZU

[Wei Ping @_weiping]

Incredible!

[Claude @claudeai]

Fable 5.2 and Mythos 5.2 are a step above Opus 5.5 in almost all the capabilities we evaluated.

Testers have found that they communicate more clearly than past Fable and Mythos models. They generate outputs up to 30% faster than Fable 5.1 and Mythos 5.1.

[Wei Ping @_weiping]

@mrtnm smart idea

[Wei Ping @_weiping]

congrats to @magicsongyang and the team!

[Nanbeige @nanbeige]

🚀Excited to announce Nanbeige4.3-33B!
To push the performance ceiling of small agents, our strategy in Nanbeige4.3 starts with building a powerful teacher first.
Nanbeige4.3-33B delivers SOTA agentic performance among open models at ~30B parameters:
* Strong results on agentic coding and cowork benchmarks.
* When integrated into mainstream agent harnesses, it tackles real-world tasks with cost-efficient inference.
* It will serve as a teacher for upcoming smaller Nanbeige models, starting with Nanbeige4.3-3B.

https://t.co/tSToQgZf2O
https://t.co/D228euvybz

#LLM #Agent

[Wei Ping @_weiping]

@LexnLin @Kimi_Moonshot https://t.co/OlYp13wGlL

[Jennifer Zhou @chem_safety]

We looked at TerminalBench 4.0 and found a lot of issues. After fixing the issues Sonnet 5.5 scores 89.2%. So, TerminalBench is already mostly saturated.

Along the way we found a lot of crazy stuff, including reward hacks in the published transcripts.

[Wei Ping @_weiping]

The new frontier open model matches the best closed models

[Kimi.ai @Kimi_Moonshot]

Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.

🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.

Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on https://t.co/85jA71X9gw, Kimi Work, Kimi Code, and the Kimi API.

Open weights by Nov 4, 2026.

🔗 API: https://t.co/XCrgjXAqMw
🔗 Tech blog: https://t.co/5iPP0MBej6

[Ming-Yu Liu @liu_mingyu]

Cosmos is a world foundation model for building your own world models. In this release, we show how to build real-time autoregressive world models for robots and autonomous vehicles. We also have a real-time transfer model to help boost realism and diversity of your favorite classical simulator outputs.

Check it out.

https://t.co/LsysE6JWk3

[Wei Ping @_weiping]

@teortaxesTex Wow, thanks

[Wei Ping @_weiping]

Ran the physics for this scene with GPT-6.1 Sol Max: lifting a 1.5-ton car’s front end could take ~686 kg-force at the bumper, far beyond ordinary human strength

[Unitree @UnitreeRobotics]

Unitree Introducing: H2-X Kung Fu Legend Guardian🤖
Powerful performance to serve and guard you, with a maximum endurance of approximately 8 hours.
Please use the robot safely and in a friendly manner.

[DeepSeek @deepseek_ai]

🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!

Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We���re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.

🔗 https://t.co/JGAik1kgk0

1/4

[Lingjie Weng @Everlynww]

What should you post?

What content actually works?

How do you show up consistently while still building your product?

For most founders and small teams, there’s simply never enough time.

So we built the marketing team we wished we had.

Meet AdAnt — your proactive, 24/7 AI social media marketing team.

Give AdAnt your product URL, and a team of AI agents gets to work:

Research your niche & competitors → spot trends → build strategies → create videos → plan your calendar → schedule & publish.

No constant prompting. No waking up wondering what to post today.

Just an AI team that keeps your social presence moving while you keep building.

AI isn’t waiting for your prompt anymore.

Today, AdAnt is open to everyone. 🚀

Your Muse for social media marketing.

[Iris Tu @iris_ye_tu]

Introducing the Muse for Social Media Marketing.

A proactive team of AI agents working 24/7 to research, create and publish your social content.

In the last 30 days alone, we’ve generated 35M+ organic views for a single account, 100M+ for us and our clients in the last 6 months.

Our content has also helped clients reduce customer acquisition costs by 60% on average when repurposed for paid ads.

Your AI marketing team is ready to work. 24/7.
- Strategist: research your niche on TikTok, Instagram, YouTube and Meta Ads.
- Creator: turns every strategy into a batch of videos.
- Analyst: listens on X and Reddit for conversations worth joining.
- Manager: assigns tasks to all agents, plans the calendar and publishes posts.

Try it for free → @adant_ai

[Wei Ping @_weiping]

@scaling01 maybe start by not reducing model capabilities to a single scalar metric?

[Wei Ping @_weiping]

meta is so back!

[Alexandr Wang @alexandr_wang]

🍉 MUSE WONDER 🍉

our research process starts with evals. we developed the Personal Superintelligence Index (PSI) internally to help us measure the important dimensions of what actually makes a model useful for personal agents, vs just focusing on external benchmarks.

the end result is that our model is incredible for personal agents in many subtle and important ways, including Active Memory and Active Focus.

i'm especially proud of our research team for taking a particular focus on developing useful models. the whole point of building advanced frontier models is to make them useful to real people, and we have thought about that every step of the way in our research process.

muse wonder is quite advanced across the board, and certainly has big model smell (which we quantified!), but is truly state of the art for personal agents.

[renjie pi @RenjiePi]

Thrilled to share that our paper has been accepted to the NeurIPS 2026 Datasets and Benchmarks Track! I feel extremely fortunate to have collaborated with such incredible co-authors and teammates throughout my first year at NVIDIA. Huge thanks to everyone involved!

[Yang Chen @ychenNLP]

Thrilled that Nemotron-Cascade has been accepted as an Oral at #NeurIPS2026! 🎉
It's our fully open-source RL recipe for LLM post-training. Huge thanks to all my amazing teammates at @NVIDIA.

model+data: https://t.co/XM7LWMz3Vd

[Wei Ping @_weiping]

Jul 31: V4-Flash-0731
Aug 13: V4-Pro-0813
Aug 21: V4-Flash-Vision-Exp
Sep 9: V4.1-Flash
you guys are insane..

[DeepSeek @deepseek_ai]

🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.

🔹 Introducing the smallest model in our new architecture family, with native visual understanding.
🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models.

1/6

[Wei Ping @_weiping]

@lintool Congrats!

[Wei Ping @_weiping]

@edliu1105 now, 4.2->4.3 😄 https://t.co/F546uLPvZV

[jietang @jietang]

Updated AA index... Now Fable 5.1 and GPT-astra have the same...

[Wei Ping @_weiping]

First time in my life seeing a benchmark scramble overnight to fit a model. It used to be benchmaxxing; now it’s model-maxxing 🤣 Build a great model, and the leaderboard will chase after you 😉

[Artificial Analysis @ArtificialAnlys]

Announcing Artificial Analysis Intelligence Index v4.2. We are accelerating elements of our upcoming v5 release with interim updates to keep pace with the frontier. Index v4.2 has more complex and realistic tasks, and more private test sets to prevent gaming

Intelligence Index v4.2 changelog:

+ AA-Briefcase, our agentic knowledge work evaluation with a private test set

+ @HelloSurgeAI's GDP.pdf, long context document reasoning across 4,592 PDF pages

- GPQA Diamond, an exceptional scientific reasoning evaluation that has now been saturated

… plus greater weighting on held-out test sets to prevent gaming, and grading infrastructure upgrades to increase robustness

This update brings the Index closer to real-world use cases with more challenging, complex and realistic tasks and private test sets to prevent gaming. We have been planning and building elements of Index v5 for months - it’s been 8 months since we launched Index v4 in January.

We have deliberately held back updates to keep the Index stable through recent major model launches. However, with the frontier moving so quickly in the past weeks, we feel it is important to deliver an immediate interim update to ensure our Index remains as relevant and useful as ever to users.

Beyond this interim update, our team is hard at work on v5 of the Index. We are planning more incremental releases in the near future. Stay tuned!

Intelligence Index v4.2 changes in detail:

➤ Adding AA-Briefcase: Our in-house evaluation with a private held-out test set, AA-Briefcase tests models on realistic agentic knowledge work tasks in complex projects built by industry experts. Models are evaluated on multi-week knowledge work projects, each with many linked tasks and thousands of input source files. AA-Briefcase combines rubric and pairwise grading to evaluate verifiable task success, analytical quality, and presentation quality, giving a holistic view of overall agentic capability in knowledge work.

➤ Adding GDP.pdf: Created by @HelloSurgeAI, GDP.pdf evaluates single-turn professional document reasoning across 100 PDFs and ten domains. Models must synthesize evidence distributed across 4,592 pages, including text, tables, charts, footnotes, and exclusions. Responses are graded against 1,275 expert-authored atomic criteria; the headline All-pass Rate credits a task only when every criterion is satisfied.

➤ Weighting to measure real-world use and prevent gaming: 40% of our Index weighting is now private, held-out test sets - double the figure from v4.1. Held-out data includes AA-Briefcase, AA-Omniscience, and solutions for CritPt. This reduces the ability for labs to game evaluations. The held-out percentage will increase further in Index v5.

➤ Improving our grading infrastructure: In AA-LCR v1.1, we have added a grading system prompt and corrected errors and ambiguities in answer keys, improving scoring accuracy. For GDPval-AA v2 and AA-Briefcase, we have improved our sampling and re-anchored the Elo scale, making ratings more stable as new models are added. For SciCode we have improved robustness of grading sandboxes to ensure slow but correct code does not count as a failure.

Key results:

➤ Anthropic and OpenAI lead the Index: Anthropic’s Claude Fable 5.1 leads the Index, followed by OpenAI’s GPT-6 Astra, which shows a 4pt gain over GPT-5.6 Sol. Meta is the third-ranked lab on the leaderboard, followed by SpaceXAI, Moonshot/Kimi, Z AI, and Google

➤ Cost per Task Pareto frontier shared by four labs: Anthropic, OpenAI, Meta and Z AI occupy the updated Cost per Task Pareto frontier

➤ GPT-6 Astra dominates the output token Pareto frontier: GPT-6 Astra is more token efficient than almost every other model near the intelligence frontier, with Claude Fable 5.1, Grok 4.5 and Gemini 3.5 Flash-Lite at either end of the curve (excludes models below 25 on the Index)

[jietang @jietang]

Thoughts About Scaling Law

Scaling, but not only of parameters. Every model release now ends with the same question: how many parameters? It isn't a question that can be answered on its own. Parameter count is only meaningful alongside three others — how much data you have, where you intend to spend your compute, and who will run the model, under what conditions.

The field learned this the hard way. Kaplan et al. (2020) fit an exponent that told everyone to grow parameters faster than data — roughly 2.7:1 — and the industry complied: GPT-3, Gopher, MT-NLG. Hoffmann et al. (2022) redid the experiment across four hundred models and found the compute-optimal split is closer to 20 tokens per parameter, and that with sufficient compute the two should grow at the same rate rather than drifting apart. The error in the earlier fit compounded with every order of magnitude of compute, which is why the largest models of that generation were the most misallocated. The trillion-parameter round was, in retrospect, a detour the whole field took together and then reversed.

Chinchilla wasn't the end either. It optimized training compute for models that would be trained once and evaluated. Today a model is called billions of times a day and inference dominates lifetime cost. Put inference into the objective and the optimum moves toward smaller models trained far longer — deliberate over-training, which is what Llama-2-7B and Gemma-2-9B were doing at roughly 290 and 889 tokens per parameter.

Sparsity moved the target again. In a MoE model two quantities have to be kept apart: total parameters govern roughly how much the model can hold — knowledge, facts, the long tail — while activated parameters and effective depth govern roughly how far it can think, how many steps of a causal chain it can carry before it comes apart. A dense 20:1 ratio does not transfer. And the ratio isn't a single number at all: Roberts et al. (2025) find the optimal tokens-per-parameter is task-dependent, with memorization favoring more parameters and reasoning favoring more data. Follow-up work on MoE observes that at fixed TPP, pushing total parameters higher actually degrades reasoning, while activating more experts reliably helps it.

This matters for what we are building toward. Finding a vulnerability is not a retrieval problem. It doesn't come from having memorized more CVEs; it comes from carrying a twenty-step chain of inference to the end without losing the thread. That capability does not live in total parameter count.

Which brings us to this release. Total parameters appear to matter up to a threshold — enough to hold the world — after which additional capability comes from scaling elsewhere: effective depth per forward pass, and above all post-training. GLM-5.3 is our controlled experiment on that claim. Same base, same architecture, same total and activated parameters as GLM-5.2. One month of scaling long-horizon environments and RL. The gains are not marginal. Well, scaling has more than one dial. We turned the post-training one this time because it had the most slack left in it — not because the others are finished. Base model size, pretraining data, compute spent per forward pass: all of them are still on the table, and we will come back to each. What this experiment taught us is that the dials do not have to be turned together, and that the one worth turning next is rarely the one that was worth turning last. We are not done scaling. Next time, maybe mid-training, pre-training, and even more.