Excited to be (finally) heading to the PyTorch Conference! I’ll be giving a talk tomorrow at 11:00 AM on “The LLM Landscape 2025”, where I’ll discuss the key components behind this year’s most prominent open-weight LLMs, and highlight a few architectural developments that go
LLMS
-

LFM2-VL-3B: New Large Vision Language Model Released
By
–
LFM2-VL-3B just dropped! It's a bigger version of our VLMs with fast inference and strong performance. Look at how gracefully it distinguishes dogs from ice cream scoops
-
Transitioning from web search to AI agent-driven browsing
By
–
This isn’t “ChatGPT in a tab.” It’s your browser as an AI agent platform. We’ve gone from “search the web” → “work with the web.” Welcome to the agentic internet. Try it now: http://
chatgpt.com/atlas -

Atlas AI adds memory features for browsing and research
By
–
2. Built-in memory Atlas remembers what you browse — totally private and under your control. Ask things like: “Summarize all the jobs I looked at last week.”
“Continue that research thread from Monday.” Delete or archive any time. -
Atlas AI assistant integrates ChatGPT directly into web browsing
By
–
1. ChatGPT, anywhere on the web No more copy-paste between ChatGPT and Chrome. Atlas understands what’s on your screen and helps you right there summarizing, explaining, researching, automating.
-
OpenAI Launches ChatGPT Atlas Web Browser Integration
By
–
RIP Perplexity Comet & all AI browsers?
— God of Prompt (@godofprompt) 22 octobre 2025
OpenAI just dropped ChatGPT Atlas a full web browser with ChatGPT built inside.
This isn’t an extension. It’s the internet, rebuilt around your AI assistant.
Here’s what it can do 👇 pic.twitter.com/VY5DLi3NuVRIP Perplexity Comet & all AI browsers? OpenAI just dropped ChatGPT Atlas a full web browser with ChatGPT built inside. This isn’t an extension. It’s the internet, rebuilt around your AI assistant. Here’s what it can do
-
GPT-10 Input Architecture: Pixels Over Tokenization
By
–
Haha. I am afraid people interpreted my “delete tokenizer” as “use bytes directly without BPE”, the issue is you *still* need bytes encoding arbitrariness even for that! Pixels is the only way. Just like humans. It is written. If GPT-10 uses utf8 at the input I will eat a shoe.
-

AI Infrastructure Transforms Finance: LLMs Power Smarter Payments
By
–
Join us at @Money2020
: AI in Finance — Scaling Financial Intelligence with @OracleCloud & NVIDIA Hear from leaders at @Mastercard
, @Stripe
, and @Fiserv on how specialized LLMs and high-performance AI infrastructure are powering smarter payments, hyper-personalization, and -
Qwen3 Small Models: Faster and Cheaper Alternative to GPT
By
–
Interesting that they mentioned faster & cheaper compared to OpenAI’s latest models not “customizable”. That makes me think they are specifically referring to gpt-oss, This in turn means they are using the small, dense Qwen3 models, maybe 0.6 to 4B range. And this is
-
Training multilingual cultural model improves multimodal benchmark performance
By
–
Thank you so much and appreciate you taking time to check the paper. We trained a model(
https://
huggingface.co/neulab/Cultura
lPangea-7B
…) on the subset of the dataset and were able to improve performance on many multilingual/cultural multimodal benchmarks. Yeahhh, would be great too if you could also
