This is one of those things that gets released with not much fanfare, but it's going to completely change the space in a few months. Running tools during the reasoning is going to be the norm.
LLMS
-
OpenAI Releases o3 and o4-mini Models
By
–
OpenAI released o3-deep-research and o4-mini-deep-research models on Playground. "These models are the same post-trained o3 and o4-mini models that power deep research in ChatGPT"
-
Gemma 3n E2B and E4B models available on AI Studio
By
–
Gemma 3n E2B and E4B models are now available on AI Studio!
-
Frontier Models Now Support Web Search in API Reasoning Flow
By
–
One thing I love about our latest frontier models is their ability to seamlessly call tools within their reasoning flow.
— Romain Huet (@romainhuet) 26 juin 2025
Web search is now supported in the API for o3, o3-pro, and o4-mini! 🌎🔎 https://t.co/IcLQB05QVvOne thing I love about our latest frontier models is their ability to seamlessly call tools within their reasoning flow. Web search is now supported in the API for o3, o3-pro, and o4-mini!
-
Gemma 3n: Powerful Open Source Multimodal Model for Edge
By
–
Our open source Gemma models are the most powerful single GPU/TPU models out there! Our latest model Gemma 3n has amazing performance, multimodal understanding, & can run with as little as 2GB of memory – perfect for edge devices – enjoy building at http://
ai.studio ! -
Google Releases Gemma 3n Open Model with Developer Guide and Competition
By
–
.
@Google Gemma 3n open model is now fully available for developers. See this guide: https://
developers.googleblog.com/en/introducing
-gemma-3n-developer-guide/
… Kaggle competition ($150,000 in total prizes): https://
kaggle.com/competitions/g
oogle-gemma-3n-hackathon
… 𝓓𝓮𝓽𝓪𝓲𝓵𝓼: What it is: The latest in the popular Gemma family of open models, but built on a -
Preference for Scalable Turn-the-Crank Algorithms with Increased Compute
By
–
I sometimes try to explain it as a statement of preference for "turn the crank" algorithms. When you're eventually given more compute (faster crank), you shouldn't have to touch anything at all, you just crank faster to make better. You can (and probably locally should) knowingly
-
GPT-4.5 Thinking Enforcement Prompt: The Underrated AI Trick
By
–
GPT-4.5 with a thinking enforcement prompt is by far the most slept-on AI trick right now. It's just so damn good.
-
LangChain Academy: Building Ambient Agents with LangGraph
By
–
Our latest LangChain Academy course – Building Ambient Agents with LangGraph – is now available!
— LangChain (@LangChain) 26 juin 2025
Most agents today handle one request at a time through chat interfaces. But as models have improved, agents can now run in the background – and take on long-running, complex tasks.… pic.twitter.com/hyaYlpstYBOur latest LangChain Academy course – Building Ambient Agents with LangGraph – is now available! Most agents today handle one request at a time through chat interfaces. But as models have improved, agents can now run in the background – and take on long-running, complex tasks.