@pronounrespectr @SouthHampUpper1 @visionergeo Retarded take. It's about wars not men's role as individual protectors.
English
Big Chungus
12.7K posts
@Ready2ESC @SouthHampUpper1 @visionergeo Gay and emotional reaction. This is a conversation about how men’s role is to protect, not women’s’
English
🇩🇪BREAKING | Starting January 1, 2026, all men aged 17 to 45 must obtain permission from a Bundeswehr career center if they plan to leave Germany for more than three months — whether for studying abroad, work, or extended travel — Berliner Zeitung.
This requirement is now in effect on a permanent basis and is no longer limited to periods of heightened tension or a state of defense, meaning a specific military threat.
See the latest updates with us: @visionergeo
English
@cryptopunk7213 Cos most of those so called "artists" are crap and they feel threatened by AI and fear they will lose their jobs. Not a baseless fear mind you.
English
genuine question: why are the arts and gaming communities so fucking touchy about ai?
i get that 99% of examples are slop but we’re reaching a point where ai-generated media is indistinguishable for 90% of the world
that shouldn’t go un-acknowledged just because you want to hide behind a professional identity
ai isn’t going away, it’s getting (a lot) better, so why not try and figure out how to work with it to your advantage? isn’t that what a bunch of hollywood is realising now?
doomer: “NO THIS IS NOWHERE CLOSE TO PIXAR GRADE”
8 year old (target audience): “haha that’s awesome”
what am i missing?
Is this a 3D model?@IsThisA3DModel
no and this is nowhere close to "Pixar-grade"
English
@pronounrespectr @SouthHampUpper1 @visionergeo As if all men were equally built. You're a retard. And go fk yourself we're not going to war to die for shitty politicians.
English
@SouthHampUpper1 @visionergeo So gay. Because women aren’t built for war you fucking freak.
English
@TheAhmadOsman Sure but not everyone wants to use the model for coding. I for example use it to analyze text in Hungarian or give analysis on various topics. For me coding and agentic tool calling are useless metrics.
English
Why Most “Local AI Setups” Miss What Actually Matters
> Single prompt demos
> no sustained load
> no concurrency
> no long context
> no KV pressure
> no batching pressure
> no scheduler stress
> no tail latency
> no failure analysis
“It runs” is not the same as
“it could manage my Agentic workflows”
Real questions are:
> What breaks first
> at what concurrency
> at what context length
> at what KV size
> under what scheduler pressure
> for how long
Loading a model,
sending one prompt,
and watching tokens come out
is NOT what you should be focused on
Ahmad@TheAhmadOsman
How Fast is Gemma 4 on a MacBook Pro M4? Benchmarking Google's new MoE (26B-A4B) > Model size: 26.1 GiB > Load time: ~4.2s Comparing single request VS > concurrent requests performance > 32k total context, 4 parallel slots single request behavior > TTFT: 5.68s > prompt: 3,701 tokens @ 652 tok/s > decode: 40.08 tok/s sequential (1 request at a time): > avg duration: 20.5s > p99: 22.1s > throughput: 40.11 tok/s > clean finishes: 100% concurrent (4 parallel requests): > aggregate throughput: 47.25 tok/s > total system throughput: 262.27 tok/s > avg duration: 65.1s > p95 latency: 68.8s > req/sec: 0.058 Head-to-Head: Sequential vs Concurrent throughput: > 40.11 tok/s → 47.25 tok/s (+17.8%) > small gain despite 4x parallelism latency per request: > 20.5s → 65.1s (~3.2x slower) > you pay heavily for concurrency system throughput (true utilization): > ~40 tok/s → 262 tok/s (~6.5x total output) > this is where concurrency wins tokens per second (decode ceiling): > ~40 tok/s steady in both modes > hardware-bound, not scheduler-bound TTFT impact: > ~5.7s baseline → buried under queueing in concurrent > “headers waittime” becomes the bottleneck What this actually means? - You don’t get linear scaling from parallel slots - You trade latency for total output - Mac Unified Memory setup is clearly saturating - Bandwidth + Scheduling overhead show up immediately This is exactly why GPUs dominate here Concurrency without killing latency
English
@VM_Falcon @Rixi__ @AbsxluteMR No, he's not. Countless times Phi's dd or torch was their win condition and he supported him. Not to mention when he flexes for support.
English
@Rixi__ @AbsxluteMR The post is about how Terra isn’t the best Because of how much his team hard focuses supporting him.
And you think sparkr isn’t the same if not more?
English
Terra praised as the best dps on the game but wait till u see how many dps could do what he did on 100t with the resources they gave him 🔥
English
@StefanFrancisci By someone who terrorized his wife? Who made secret recordings of her and started his political career by releasing those? The one who only worked his entire adult life in NER? The one who was celebrating Orban's victory in 2022? The one who was partying in drug fueled parties?
English
@Ljt019117161 @liyucheng_2 But I said at least as good or a bit better (122B that is) and your benchmarks support this too isn't it? I mean 122B beat 27B in some benchmarks by a bit and lost in a few by a bit. Even though it is a lot bigger. (Make no mistake I prefer the 122B but still)
English
@Ready2ESC @liyucheng_2 I’m kinda annoyed this doesn’t have coding on it tho, I wonder if it’s grouped under STEM or smth
English
Folks, pls vote Qwen3.6 27B
Chujie Zheng@ChujieZheng
We are planning to open-source the Qwen3.6 models (particularly medium-sized versions) to facilitate local deployment and customization for developers. Please vote for the model size you are **most** anticipating—the community’s voice is vital to us!
English
April Model Releases (So Far)
- Meta Avacado
- Deepseek V4
- GPT-5.5 (“Spud”)
- Gemma 4 series✅
- Qwen3.6-Plus✅
- Qwen3.5 Max Pro✅
- Qwen3.5 Omni Plus✅
- Gemini 3.1 Flash Live ✅
- GLM-5.1 open weights
- GLM-5V-Turbo✅
- MiniMax M2.7 open weights
- MiniMax M3.0
- Kimi K3.0
- Claude 5 (“Mythos”)
- StepFun
- Hunyuan 30B MoE
- Trinity Large Thinking (🇺🇸) ✅
- 1-bit Bonsai 8B (🇺🇸) ✅
- Holo3 (🇫🇷) ✅
English
@kurk165 @GrandMasterV @ChedarFederline @ricwe123 No I'm not calling you worthless for whatever you belive in. I call you worthless because you literally are here: you say nothing of value, zero facts zero evidence only trolling. That is worth nothing.
English
@Ready2ESC @GrandMasterV @ChedarFederline @ricwe123 Thank you! It's an honor to be called worthless by a Putinist. Then I know we don't share the same values.
English
Back in 2014 Ukrainian soldiers were shooting into the homes of civilians in East Ukraine.
Their mission was to spread terror and crush any dissent against the CIA backed coup in Kyiv.
This is something the western mainstream media should be telling you.
But they won't......
English
@kurk165 @GrandMasterV @ChedarFederline @ricwe123 You are literally worthless and provide nothing of substance other than your trolling.
English
@GrandMasterV @ChedarFederline @Ready2ESC @ricwe123 Putler is to blame. Without his megalomania, there would have been no war. To save Russian and Ukrainian lives in the future, he can withdraw his poor soldiers from Ukraine and peace will prevail.
English
English
@Ready2ESC @ricwe123 It is known that Putler is a war criminal. Do you believe me now?
English
@liyucheng_2 That is what I noticed, at least in Hungarian. 27B makes a lot more mistakes and forms unnatural sentences compared to 122B.
English
@Ready2ESC Valuable feedback. So larger MoEs may actually do better in multilingual?
English
@Ljt019117161 @liyucheng_2 Active parameters are only 10B so often times the dense model with all its 27B parameters active outperform the bigger MoE model in tasks like coding and agentic tool usage. The 122B has the bigger knowledge but if the router doesn't select the relevant experts it can be worse.
English
@Ready2ESC @liyucheng_2 It’d be more surprising if it wasn’t better wtf it’s 5x as many params even if it is a moe
English
@ChujieZheng Please release models in the range of 122b, that is a sweat spot, bigger than that is prohibitive to run, smaller than that is either worse in quality or very slow because it has to be so dense.
English
We are planning to open-source the Qwen3.6 models (particularly medium-sized versions) to facilitate local deployment and customization for developers. Please vote for the model size you are **most** anticipating—the community’s voice is vital to us!
English
@KogdaVojne @Lana1142086 @NullusPoetry He's way better than anyone from the opposition. The only thing holding the opposition together is their hatred. Nothing more.
English
@Ready2ESC @Lana1142086 @NullusPoetry Unlikely, because he's a shit leader who can't govern the country. His last 4 years were an era of hyperinflation, obscene levels of corruption, replacement migration and utterly deranged, retarded plans of re-industrialization. He's a shit leader who will fall because he's shit.
English
Pay close attention to the language, especially the phrase, “under normal democratic conditions.”
These polls and this framing are a blatant psyop, designed to gaslight and emotionally charge voters.
This is how immoral, unscrupulous actors engineer narratives.
The shills, glowies, and operatives are already breaking ground for a new Maidan.
Szabolcs Panyi@panyiszabolcs
💥Fresh polls ahead of Hungary’s April 12 parliamentary elections show the opposition TISZA party’s lead over Viktor Orbán’s Fidesz is holding or widening. Under normal democratic conditions, it’s hard to see Orbán closing such a large gap. But this campaign isn't normal at all.
English
@RealAraZobayan @Fredcel193345 @rustycheems @Lana1142086 @NullusPoetry No but really anyone sane in Hungary knows nothing is decided yet and it is very close.
English
@Fredcel193345 @rustycheems @Ready2ESC @Lana1142086 @NullusPoetry So true! Either he wins, or he doesn’t
English