Learn how to build with OpenAI models and AI agents in my new courses here: https://dair-ai.thinkific.com/
Use code CM24 to get a 35% off. The offer expires in 3 days!
—
Overview of OpenAI full o1 model inside of ChatGPT.
00:00 Overview
02:27 Inference Speed
05:38 Math Puzzle
06:28 Math Reasoning
08:11 Complex Knowledge Understanding
11:46 Image Understanding
14:13 Coding
#ai #chatgpt #artificialintelligence
source
Learn how to build with OpenAI models and AI agents in my new courses here: https://dair-ai.thinkific.com/
Use code CM24 to get a 35% off. The offer expires in 3 days!
I have had mixed results personally. I found o1-preview did not reliably produce better results. It frequently brushed over implementation details and was unwilling to get in to deep and complex topics.
It's like when they released 4 and it was amazing…then it immediately got watered down
Since when does it start to cost 200 dollars, I pay 20 dollars
Holy everyone has been giving their opinion on how bad this new o1 is when it comes to giving accurate answers, o1 preview wins everything (except for the speed of responding huh it can accurately solve a hard math problems
o1 gives so short answers, i liked the verbose response from o1-preview much better. also it was thinking longer and better.
no way that a model so fast is better. They gave us again some watered down model and internally and to business they give the state of the art model. Also seeing in your video that it cant even solve basic word puzzles, everybody can think what model we have here.
In my experience o1 is actually way worse than o1-preview at coding anything beyond asking it a very simple logic questions
super cool!!I'm increasingly excited to see what surprises the next model will bring us!
This is a very step forward towards more reliability for a model.
Thanks for sharing your experience. I've found that the o1 model with its response rate and additional intro text to be less performant than Anthropic's Claude-3-5-sonnet in this use case. As soon as I submitted my question, I got a response within the second. Additionally, I got what I wanted without the additional into text that would have cost me 10x or more with o1 API pricing when it comes out.
Granted, it does have more reasoning and evaluating capabilities but I mostly see that in coding. However, the costs eclipse the benefit in this use case.
SEE – no intro text and response within a second. This o1 hype needs better use cases to demonstrate why I should use it instead of Sonnet-3-5. Great job reporting, though.
—————————————————————————————————-
Here are the major Roman emperors of the 2nd century CE:
Trajan (98-117 CE)
Expanded empire to its greatest territorial extent
Extensive building program including Trajan's Forum and Column
Improved infrastructure with roads and harbors
Instituted social welfare programs
Considered one of the "Five Good Emperors"
Hadrian (117-138 CE)
Built Hadrian's Wall in Britain
Consolidated rather than expanded empire
Extensive building projects including Pantheon reconstruction
Traveled widely throughout the empire
Known for cultural achievements and philhellenism
Antoninus Pius (138-161 CE)
Peaceful reign with economic prosperity
Maintained borders without major wars
Built temples and public works
Strong financial management
Known for just and moderate rule
Marcus Aurelius (161-180 CE)
Philosopher emperor who wrote "Meditations"
Defended empire against Germanic tribes
Co-ruled with Lucius Verus (161-169)
Dealt with Antonine Plague
Last of the "Five Good Emperors"
Commodus (180-192 CE)
Son of Marcus Aurelius
Marked end of "Golden Age"
Known for megalomania and excess
Participated personally in gladiatorial games
Assassinated in 192 CE
You're comparing to open source (local-run you meant I think) models but ChatGPT has incomparable hardware access