Forget GPT-5… Anthropic’s Sonnet 4.5 Just Changed Everything

1. 

   @TheAiGrid

   November 16, 2025 at 3:37 am

   Want to stay up to date with ai news for free? – https://aigrid.beehiiv.com/subscribe

2. 

   @DenofLore

   November 16, 2025 at 3:37 am

   Ech Its ok. A good teammmber in addition to but not a replacemnt for.

3. 

   @greatwazzoo

   November 16, 2025 at 3:37 am

   Earth shattering Gemini update in 3,… 2,… 1,…

4. 

   @mrd6869

   November 16, 2025 at 3:37 am

   It looks decent so far.
   Im also waiting for the response from china open source as well as Google/openAI

5. 

   @g.orbeliani.1

   November 16, 2025 at 3:37 am

   ahahahah world just changed again lolol

6. 

   @DynamicUnreal

   November 16, 2025 at 3:37 am

   I dislike Claude 4.5 mainly because it doesn’t have any sycophancy. I mostly use LLMs to conversarte with them about my life, mainly bounce thoughts, events, situations and possible solutions. This latest version of Clause is a moralizer. “Hey Claude today I did XYZ “and then it says “what you did was wrong and unethical.” It is rigid in its PC ideological thinking.

7. 

   @animeplustrance

   November 16, 2025 at 3:37 am

   How come it wasn't entered in the coder olympiad?

8. 

   @quickmythril2398

   November 16, 2025 at 3:37 am

   i don't understand the meaning behind the 7 hr vs 30 hr benchmarks. i thought the main limitation was the context window size. what does it mean to be able to code autonomously for that long? why does it stop at all? do they have some hardcoded time limit where they need to take a break? what exactly was changed that allowed them to see that increase 4x in 4 months?

9. 

   @Aaron_Mottram

   November 16, 2025 at 3:37 am

   Is it only good at coding?

10. 

    @annonymbruger

    November 16, 2025 at 3:37 am

    Day after day new ai generated analyses from so many using same context same result all over day after day. This is turning into a nightmare 😢

11. 

    @shd94

    November 16, 2025 at 3:37 am

    🥱🥱🥱 It's a pity this kind of excessive exuberance is widespread around AI. Yes, it's great, but not everything is a breakthrough. Tomorrow, you'll come up with another game changer.

12. 

    @adamgibbons4262

    November 16, 2025 at 3:37 am

    If it gets to 90%+ on SWE bench, then why not give it its own source code to improve?

Comments are closed.