Zhen Dong

Incoming Assistant Professor @UCSB, AI Researcher @NVIDIA, PhD/Postdoc @berkeley_ai

[OpenAI @OpenAI]

On Agents' Last Exam, GPT‑5.6 Sol sets a new high of 53.6, eclipsing Claude Fable 5 (adaptive) by 13.1 points.

At medium reasoning, it beats Fable 5 by 11.4 points at roughly one-quarter the estimated cost. GPT‑5.6 Terra and Luna also outperforms Fable 5 at around one-sixteenth the cost.

[Yiyou Sun @YiyouSun]

“AI agents will outperform humans at almost all jobs by 2026–2027.” - The forecast is everywhere.
So we built the exam to test that claim, on real labor-market aligned work. On the hardest tier, top agents pass 2.6%.
Meet Agents' Last Exam (ALE), a rolling benchmark measuring whether agents can actually do real jobs. 🧵👇

[NVIDIA @nvidia]

Introducing NVIDIA Nemotron 3 Ultra.

A frontier smart open model built for long-running agents that need to plan, reason, use tools and keep working across complex coding, research and enterprise workflows.

Up to 5x faster inference and up to 30% lower cost for agentic tasks.

Learn more: https://t.co/h9XLqqYPFf

[NVIDIA @nvidia]

https://t.co/8zNQX2OSSF

[DINQ @dinq_me]

You’ve done real work.
But most of it is hard to see.

DINQ brings your projects, code, and research onto one card.
No self-promotion. Just real signals.

Build your DINQ → https://t.co/IaEtN0Vkab

#DINQ

[Zhen Dong @zhendongucb]

Have a look at our newly open-sourced model Nemotron 3 Nano 30B-A3B. We achieved AA Index 52.
The efficiency from MoE Hybrid Mamba-Transformer model offers up to 3.3× higher throughput versus similar sized models.

[Artificial Analysis @ArtificialAnlys]

NVIDIA has just released Nemotron 3 Nano, a ~30B MoE model that scores 52 on the Artificial Analysis Intelligence Index with just ~3B active parameters

Hybrid Mamba-Transformer architecture: Nemotron 3 Nano combines the hybrid Mamba-Transformer approach @NVIDIAAI has used on previous Nemotron models with a moderate-sparsity MoE architecture, enabling highly efficient inference, particularly at longer sequence lengths

Small-model improvements: with 31.6B total and 3.6B active parameters, Nemotron 3 Nano scores 52 on our Intelligence Index, in line with OpenAI’s gpt-oss-20b (high). This represents a +6 point lead on the similarly-sized Qwen3 30B A3B 2507 and +15 improvement on NVIDIA’s previous Nemotron Nano 9B V2 (a dense model)

High openness: Nemotron 3 Nano follows other recent NVIDIA models in open licensing and releases of data and methodology for the community to use and replicate - it scores an 67 on the Artificial Analysis Openness Index, in line with previous Nemotron Nano models

Key model details:
➤ 1 million token context window, with text only support

➤ Supports reasoning and non-reasoning modes

➤ Released under the NVIDIA Open Model License; the model is freely available for commercial use or training of derivative models

➤ On launch, the model is being made available with a range of serverless inference providers including @baseten, @DeepInfra, @FireworksAI_HQ, @togethercompute and @friendliai, and it is available now on Hugging Face for local inference or self-deployment

See below for our full analysis and key announcement links from NVIDIA 👇

[Zhen Dong @zhendongucb]

Our team at NVIDIA has openings for full-time roles. If you are interested in LLM/DLM Post-training and AI Systems, please feel free to contact us.

[Jiantao Jiao @JiantaoJ]

🚀 We’re hiring at NVIDIA!
Our team is pushing the frontier of LLM / DLM post-training and system optimization. We are looking for exceptional people with large-scale LLM + systems experience to join us (full time only).

🔹 Focus areas include:

•Post-training of large models

•Systems for LLM/DLM training & inference at scale

•Efficiency, scaling, and evaluation frameworks of LLMs

At NVIDIA, you’ll work with world-class researchers and engineers on cutting-edge foundation models at unprecedented scale.

👉 If you’re passionate about LLMs, systems, and building the next generation of AI, we’d love to hear from you.

📩 If you’re interested, please send me your CV!

@nvidia #LLM #AI #Systems #PostTraining #DeepLearning

[Zhen Dong @zhendongucb]

Nano-V2 is a 9B model pretrained from scratch that can outperform Qwen3 8B. We are open-sourcing both the models as well as the pretraining dataset.
Please check our Tech Report for more details: https://t.co/2Oq35q8ssR

[Oleksii Kuchaiev @kuchaev]

We are excited to release Nvidia-Nemotron-Nano-V2 model! This is a 9B hybrid SSM model with open base model and training data. This model also supports runtime "thinking" budget control. HF collection with base and post trained models: https://t.co/n3M01d8lSm

[NVIDIA AI Developer @NVIDIAAIDev]

We just released 3 million samples of high quality vision language model training dataset for use cases such as:

📄 optical character recognition (OCR)
📊 visual question answering (VQA)
📝 captioning

🤗 Learn more: https://t.co/zUEiB6ZLZR
📥 Download: https://t.co/JD7dZ73oA6

[NVIDIA AI Developer @NVIDIAAIDev]

🏎️ ️We accelerated @OpenAI’s new GPT open weight models -- gpt-oss-20b and gpt-oss-120b -- for leading inference performance on NVIDIA Blackwell architecture, delivering up to 1.5 million tokens per second on an NVIDIA GB200 NVL72 system.���️🏁

👀 Technical deep dive on how we maximized model performance with the OpenAI community including @huggingface, @ollama, and @vllm_project so you can deploy them from the Cloud to the Edge.

Tech Blog ➡️ https://t.co/HHChnErTJT

[Zhen Dong @zhendongucb]

Excited to share that our Super-V1.5-49B achieved a great Artificial Analysis score of 64, with a significant +23 points gain from redoing post-training, compared to its base model llama-3.3-70B-instruct (score 41).

[Artificial Analysis @ArtificialAnlys]

NVIDIA has released the latest member of its Nemotron language model family, Llama Nemotron Super (49B) v1.5, reaching a score of 64 on the Artificial Analysis Intelligence Index.

The model is an evolution of Super 49B v1 from earlier this year, with advances from post-training on new reasoning datasets generating a 13-point increase in the Intelligence Index. This puts @NVIDIA’s latest Super 49B release ahead of their previous Ultra 253B parameter model, despite having less than 1/4 the parameters.

Leading dense model performance: with this latest iteration, Nemotron Super 49B v1.5 is the only dense model in the top 5 open weights models, competitive with much larger recent MoEs from Alibaba, Deepseek and MiniMax.

Key model details:
➤ Retains the same 131k context window as Nemotron Super v1
➤ Supports reasoning or non-reasoning modes with ‘/no_think’ settings in the system prompt
➤ Released under the NVIDIA Open Model License, as with previous Nemotron models

[Oleksii Kuchaiev @kuchaev]

Llama-Nemotron-Super-V1.5 got AA intelligence index of 64. This is more than previous Ultra (61) model and is by far the "smartest" open-weights dense model. The best model for deployment on a single H100. Head to @ArtificialAnlys for detailed analysis https://t.co/DB0zzky62h

[Zhen Dong @zhendongucb]

Meet our new 49B open-sourced model Llama-Nemotron-Super-V1.5 that can outperform previous 253B model with only post-training🚀

[Oleksii Kuchaiev @kuchaev]

Very excited to announce Llama-Nemotron-Super-V1.5! Super-V1.5 is now better than Ultra-V1. This is currently the best model that can be deployed on a single H100. Reasoning On/Off and drop in replacement for V1. Open-weight, code and data on HF https://t.co/bePZEQJllC

[NVIDIA AI Developer @NVIDIAAIDev]

📣 Announcing Llama Nemotron Super v1.5 📣

This release pushes the boundaries of reasoning model capabilities at the weight class of the model and is ready to power agentic applications from individual developers, all the way to enterprise applications.

📈 The Llama Nemotron Super v1.5 achieves leading reasoning accuracies for science, math, code, and agentic tasks while delivering up to 3x higher throughput.

Try it on https://t.co/Y4hX79bVrQ, or download from @huggingface: 🤗 https://t.co/K6e2Zwk1D3

Tech blog: https://t.co/5gFnNM0R8m

[Igor Gitman @igtmn]

We've released a series of OpenReasoning-Nemotron models (1.5B, 7B, 14B and 32B) that set new SOTA on a wide range of reasoning benchmarks across open-weight models of corresponding size.

The models are based on Qwen2.5 architecture and are trained with SFT on the data generated with DeepSeek-R1-0528.

A few highlights 🧵

[Zhen Dong @zhendongucb]

Introducing our new work DrafterBench, evaluating the proficiency of LLMs in Civil Engineering tasks.
DrafterBench provides a technical platform to test abilities such as structured data understanding, function execution, instruction following, and reasoning.
Paper: https://t.co/oJgqzZKlrc
GitHub: https://t.co/lnrHJ4go2f
HuggingFace: https://t.co/XLXFisVw82