The diff is mostly the inference time config.
@yitayml
-
Gemini Deep Think Model Launches with IMO Gold Medal Achievement
By
–
The Gemini Deep Think model that achieved IMO gold medal 🥇 is launched! This is a general purpose model that is not only SOTA at math/proofs but also reasoning, code and many others! 🔥
— Yi Tay (@YiTayML) 1 août 2025
The exact config that achieved IMO gold with scaled up Deep Think is being made available… https://t.co/7Ol5iZQN3lThe Gemini Deep Think model that achieved IMO gold medal is launched! This is a general purpose model that is not only SOTA at math/proofs but also reasoning, code and many others! The exact config that achieved IMO gold with scaled up Deep Think is being made available
-
Architecture Ablation Study and LLM Objective Functions Analysis
By
–
It clearly ablates the main architectures like encoder decoder, decoder only etc. it even tried prefix lm. Also a master class in llm objective functions.
-
T5 Paper Remains the Most Insightful LLM Pretraining Classic
By
–
I think t5 paper is still the best classic most insightful paper for LLM pretraining. Way ahead of it's time.
-
Training a Strong Model with Trust and Morale Support
By
–
Had a super fun time training this model. A big yolo run that resulted in a super strong model. Most important thing is to trust your model and give it morale support. Was also a big eye opener to see how prep for IMO is done. Before this I knew absolutely zero about this
-
AI Progress: Imperceptible Daily Changes, Remarkable Long-term Growth
By
–
Watching AI progress is like watching your child grow up. You don't notice the little day to day changes but all of a sudden after some period of time you look back and you become amazed and shocked by how much have changed. Every minor revision to the model, every
-
Yitay expresses pride receiving compliment from renowned AI researcher
By
–
Thanks swaroop. Feels so proud to get complimented by the coolest ai researcher!!
-
Model Config Named After Cat Figure on Desk
By
–
Another fun fact that this model config was named "imocat". I was thinking of a unique name before launching the job for the first time and saw this on my desk.
-
In-context learning and competitive performance in systems
By
–
+1 Two points
1) this is really just good old in context learning
2) as already mentioned, another system without this also scored gold.