scumfuck flowergirl // i know a little bit about computers // priv: @Zbuildstd

17
Joined December 2020
assume a hypothetical LLM can fit into whatever hardware configuration you have. what is the minimum equivalent capabilities you’d cancel your frontier subs for, to use the model locally permanently?
12%opus 4.8 equivalent
6%gpt-5.5 equivalent
66%opus 5.5 equivalent
15%gpt-6-astra equivalent
3,741 votes • 16 hours
64
2
159
14,083
clarification because people are misinterpreting: - the model will prefill and run as fast as qwen3.8 27b does on your hardware - this isn't a trick question about never being able to use another model i'm just curious as to what sort of capability would cause people to unsub
3
19
1,158
nullptr 🐱🍩 retweeted
Mobile dev just moved to the cloud. Capy can now build and test your iOS and Android apps on cloud phones, powered by @tryrevyl
36
23
17
223
19,206
i think it takes genuine skill to make a 218B model perform so terribly whilst still being coherent
In case it wasn't obvious enough from the sloppiness of the report, Cohere's business involves selling their third rate LLM as part of a self-hosted offering for tech illiterate boomers in government/industry who are paranoid about Chinese models.
10
16
1
761
20,014
nullptr 🐱🍩 retweeted
found a neuron in qwen3.5-4b that makes it talk like her
84
934
111
12,511
504,417
nullptr 🐱🍩 retweeted
at walmart rn and everyone looks abliterated and quantized
30
258
12
4,279
82,335
"guys, anthropic is thrashing us in everything! we're not even top FIVE on artificial analysis! wait, i know what to do... release a model only as good as mimo-v2.6-pro, for over double the price per task! yeah!"
GPT-6.1 Terra: delivering fast and capable computer use at a lower price. GPT-6.1 Terra understands what’s on screen, takes action across apps, writes naturally, and is a more capable model for your whole workday with the efficiency to take on more of it.
14
13
1
547
17,977
nullptr 🐱🍩 retweeted
180
1,489
238
13,249
173,764
the first *good* open-weights full duplex LLM?! hope it fits on 1x4090...
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence. block.xyz/dessa
5
90
3,633
is there like, genuinely no good way to use kimi models? they're so large and expensive it's almost never worth using them at API pricing, but the coding sub moonshot provides just isn't good either
5
86
1,900
nullptr 🐱🍩 retweeted
the fuck you smiling about bitch
44
1,499
31
18,820
141,019
nullptr 🐱🍩 retweeted
i almost died in a horrific microsoft authenticator accident
21
4,076
49
35,851
362,399
nullptr 🐱🍩 retweeted
IS THE WEBSITE DOWN FOR ANYONE ELSE?
19
99
11
753
22,432
nullptr 🐱🍩 retweeted
154
528
207
5,739
1,623,954
there is not a response in the world that could possibly justify having “4” as a private key for an intermediary CA
Replying to @GoI_MeitY
However, some social media accounts are indulging in fear mongering about the impact of this incident. The facts of the case are as follows: eMudhra operates in two distinct areas:
4
30
800
7,816
nullptr 🐱🍩 retweeted
2
1
38
787
railway is unusable because of this. i wasn’t going to switch away because of their excellent service as of late but this might just be the final straw…
do i know anyone at @Railway?? just found a P0 bug on their dashboard
1
36
843
so that was surely one of the most eventful days in tech, right
15
6
231
4,230
do you guys think they'll make 4 an illegal number now
2
4
153
1,152