le premier modèle à avoir atteint le 120k parfaitement Pour l'instant, tout ce que je vois de o3, c'est un monstre
LLMS
-
ChatGPT 4o with Built-in Image Generation Capabilities
By
–
ChatGPT; was apparently 4o and the Image Generation that now comes out of the box. Prompt too boring to be worth mentioning; shorter than original tweet.
-
Multimodal AI debugging code to solve a maze
By
–
Yes; heavy tool use. It seems to have mostly solved it via code (PIL, cv2), but using multimodal intuition to debug. E.g. one attempt within the CoT generates a path that simply traverses the outside of the maze, but it recognizes on its own this is wrong so it refines the code.
-
ChatGPT solves a 200×200 maze in one try
By
–
ChatGPT o3 found a path through this 200×200 maze for me in one try. I had to overlay the solution over the original in Photoshop and flip between layers while zoomed in to check the solution never crosses a wall and none of the walls are changed. It's perfect.
-
Chatbot Arena transforms from research project into commercial company
By
–
this was a fun one: Chatbot Arena (aka @lmarena_ai
) is turning into a real company — the people behind it want to grow the popular chatbot ranking website from a research project to a business. -
Building AI Models for Better Enterprise Customer Experience
By
–
What are the advantages of building models and products? A better, more customizable experience for customers across the stack.
— Cohere (@cohere) 17 avril 2025
CEO @aidangomez joins @jacobeffron on @Redpoint's Unsupervised Learning Podcast for a wide-ranging conversation on model architectures, enterprise… pic.twitter.com/HEumCBfqvzWhat are the advantages of building models and products? A better, more customizable experience for customers across the stack. CEO @aidangomez joins @jacobeffron on @Redpoint
's Unsupervised Learning Podcast for a wide-ranging conversation on model architectures, enterprise -
Token Usage Exceeds Pro Plan Monthly Limit
By
–
I'm definitely using more than Pro's $200 per month in tokens.
-
OpenAI o3 Model Reaches Near-Genius Level Intelligence
By
–
Comme je l’ai affirmé, le modèle OpenAI o3 est à un niveau proche – voire équivalent – à celui d’un génie. Je suis sûr que certains vont chercher à se rassurer en disant : « Oui mais il ne sait toujours pas faire ci ou ça… » Ce genre de remarque est assez absurde quand on
-
Comparing Gemini 2.5 Pro vs GPT-4 Mini for Coding Tasks
By
–
Tout dépend de ce que tu veux faire. Pour le code, j'étais sur Gemini 2.5 Pro depuis une semaine car il est imbattable. Mais GPT-3.5 et GPT-4 Mini ont l'air monstrueux également. C’est sorti hier, donc je ne les ai pas encore testés assez pour savoir ce qui vaut vraiment le coup.
-
Cost-per-solution vs cost-per-token for AI models
By
–
I think what Aiden’s saying is that cost per solution makes sense but cost per token does not. If a model A gets the right answer as reliably as model B but at lower cost, the number of tokens it generated to achieve that doesn’t matter.