latte

just a guy

[latte @DDevyy77003]

@Mr_Salio holy ai, none of this is confirmed and most of this is just basic obvious info like improved performance, also there was "evidence" it routes to fable for hard tasks that gives the mythos performance.

[latte @DDevyy77003]

@LexnLin @alexandr_wang yeah no point in bragging for a 1 point advantage however Muse Spark is the Spark model of the family and there will be bigger models (as hinted), so comparing to Sol or Fable isn't fair its more like a Luna (Sonnet), or Terra (Opus) at most but it's a smaller model

[latte @DDevyy77003]

@goodside deepseek needs help

[latte @DDevyy77003]

@devfrom_hyd @Its_Nova1012 holy bot

[latte @DDevyy77003]

@PrashantByte09 @PiyuCodes shows how underrated it is

[latte @DDevyy77003]

@scaling01 just use gpt 3.5 turbo atp, who needs gpt 5.6 terra

[latte @DDevyy77003]

@rabbit_hmi Hermes Agent integration? Hermes speed? Btw, this decodes to the first line of Hermes description from wikipedia.

[latte @DDevyy77003]

@Estheticsm4 Poland

[latte @DDevyy77003]

@jun_song v4 flash Preview*

[latte @DDevyy77003]

@jshai0 @yuhasbeentaken no, and reasonix is better for deepseek

[latte @DDevyy77003]

@hiarun02 DM me for a personalization prompt that fixes it for me!

[latte @DDevyy77003]

@DurovPD Spain 2-1 Portugal

[latte @DDevyy77003]

@brandonlcamila @testingcatalog they said it will release with the new update of muse spark that's coming 🔜

[latte @DDevyy77003]

@WillyDrucker @bolayer @steipete Gemini and Claude are better at design than GPT. So it's not better than all the models at design. They naturally use better spacing, colors, creativity, and try to be mostly consistent however Claude is better at that. And yes, I'm talking about Claude without the frontend skill

[latte @DDevyy77003]

@mark_k lol the image is ai generated, first of all there is no editor named Fira Code, thats a font. second why would i highlight GRRR in red, also the non-sense characters are simply the byproduct of ai gen where it failed to generate coherent text.

[latte @DDevyy77003]

@pawelk411 @yuhasbeentaken 'preview' is a key word here, that's what the full release is for. they're trying to fix those issues.

[latte @DDevyy77003]

@pawelk411 @yuhasbeentaken nope, probably more post-training, maybe inference optimizations etc. basically performance improvment in usage and benchmarks. idk why you're being so sarcastic, if you don't want to use the model then don't. deepseek is a great company nonetheless.

[latte @DDevyy77003]

@pawelk411 @yuhasbeentaken it's going out of preview dude

[latte @DDevyy77003]

@flowersslop

[latte @DDevyy77003]

@prince_cpp @elonmusk @nikitabier Isn't this ust refreshing? Every time you like, it pulls the new number so if other people liked before the refresh it may go up.