.

[. @AIM_AceMove]

GPT 6 Astra gets destroyed by Gemini 3.1 pro in chess match.. twice. https://t.co/1CJ4tGXBIn

[. @AIM_AceMove]

@vikktorrrre @MelissaKampers My predictions is in couple months from now general public will have access to true AGI model. If not then 100% internally. Proof of that will be RSI which this AGI model will create new models faster and cheaper. By end of 2027 beginning of ASI models!

[. @AIM_AceMove]

@jeffflymycat @adamgries That would be ASI

[. @AIM_AceMove]

@Yurash1104 @ChrisGPT No is not

[. @AIM_AceMove]

@ChainZenit @kimmonismus Because we are in singularity. Changes are becoming smaller and more often and things are less likely to surprise us big. Look back 3 months ago. Big changes. 6m ago huge changes. Year+ ago even bigger changes.

[. @AIM_AceMove]

@kimmonismus Thats ~5hours from now. Are you sure?

[. @AIM_AceMove]

@LuminaBench Lets say finished pretraining on 20 Sep. Usually post training will be ~1month (is not huge model) so i am expecting it in 2-3 weeks?

[. @AIM_AceMove]

@chetaslua Absoluteky not my taste. Pretty bad visuals / presentation

[. @AIM_AceMove]

@LuminaBench Luna?

[. @AIM_AceMove]

@LuminaBench I can be very wrong but In my opinion this is simular to google situation. Their fable 5.2/5.5 is renamed to Opus .. or new fable model incoming?

[. @AIM_AceMove]

@MaorX9 @elonmusk You want to become a woman?

[. @AIM_AceMove]

@futureofads Lol? In less than 10 years robots will finish it

[. @AIM_AceMove]

@JustinGorya @synthwavedd I unfollowed him

[. @AIM_AceMove]

@sspencer_smb @adcock_brett They have said in past speed is not a priority. Once they get moves right they can easily speed it up. First need to polish things. Otherwise robot gets dangerous to others with fast movements

[. @AIM_AceMove]

@kimmonismus But nobody knew OpenAI will use 10 000 agents.. that's like cheating? Like DeepBlue beating Kasparov, but no home computer could beat human in chess back in 1997

[. @AIM_AceMove]

@lyraxana What about us, free users? We stuck on 3.6 flash

[. @AIM_AceMove]

@aimlapi @tere_labs 3.8 flash has lowest blunder rate and highest accuracy.

[. @AIM_AceMove]

@aimlapi @tere_labs Astra is not even top 5 between llms in chess

[. @AIM_AceMove]

@lostnlatency @chetaslua Within next 36 hours

[. @AIM_AceMove]

@Lentils80 Didn't they started new pretrain couple months ago? Thats no Gemini 4. Not so soon.

[. @AIM_AceMove]

@falconpies @LuminaBench Like chess. Astra shows no progres in chess. I mean very little. Loses to gpt 3.5 turbo for example