@Kayoshi666
Joined May 2026
  • Tweets115
  • Following10
  • Followers3
13Media
DeepSeek V4.1: A 0.1 update with a massive hidden catch. Going from V4 to V4.1 looks like a minor version bump. But look under the hood: they scaled the parameters from ~100B up to a 552B MoE. 🤯 🧵👇 Thread below:
1
34
Here is the brutal reality: simply increasing model size isn't real AI progress. It's brute force. They've always had the hardware (GPUs) to do this. But raw compute isn't innovation. So why do it now?
1
16
DeepSeek originally wanted to win via clever algorithms. But once they hit a wall, they just inflated the volume and tweaked the architecture to fit the expansion. Algorithmic elegance vs. raw compute. Did they take the easy way out? 🧐👇 #AI #DeepSeek #LLM
11
Hear me out |👀 It’s too early to crown GPT-6 Astra over Grok. Wait for Grok 4.7 first — that’s when this actually becomes a real fight ⚔️ Not “pretty close.” A real contender. xAI didn’t ease into the race. They showed up late and started sprinting 🚀 #Grok vs #OpenAI
24
Who's better?
100%GPT 6-astra
0%Fable 5.1
3 votes • Final results
25
Here it is — superintelligence: GPT‑6 Astra. 🤯 The benchmark doesn’t lie. We’ve been rewarded. We actually got the best one. The future isn’t approaching — it’s already breathing down our necks. ⚡ #OpenAI
36
Just saw the benchmark. Oh yes — our time has come. We actually got the best model. The future isn’t coming. It’s already here. #OpenAI
27
OpenAI I’m on my knees.When is GPT-6 Astra actually dropping??At least let a broke person who can’t even afford Plus try it for five minutes.Or just show a 10-second teaser. I’m not asking for much.
9
#AIMuse Spark 1.3 is out. Stronger on coding and agents, cheaper than GPT 5.6 Sol and Opus 5. Live in Muse Code.Who’s trying it?@AIatMeta
2
52
Why do we need Gemini 3.8 when we have Grok 4.6? It looks more expensive and slower… but it’s not. Same cost per task. Fewer tokens. Fewer steps. Speed ≠ better. Quality + efficiency matter more. Gemini = fast assistant. Grok = gets the job done right.
22
GLM-5.3 vs Gemini 3.8 Flash Same prompt: “Create a 3D shooter game using only HTML” Results: • Gemini — 33.9s / 795 lines • GLM — ~6 min / 1058 lines Gemini is fast. But what’s the point if GLM does it right on the first try? Pure one-shot. No retries. #GEMINI #GLM Videos 👇
1
2
197
OMG, Gemini 3.8 Flash benchmarks just dropped. Cheap, fast, and surprisingly strong in some areas. Claude still wins the agent game though. Main screen runs on Fable 5.1 – and now we know why. #Gemini3.8 #AI #Benchmarks
1
2
1,292
Just glanced at the latest benchmarks. Fable 5.1 doesn't just beat 5.0—it's smarter and cheaper. Upgrade unlocked. 🚀
25
Dude, I love AI so much… but all I can do is watch from the sidelines. I’m broke — can’t afford a single subscription, and there are no decent free unlimited options left. It’s pure torture watching everyone else live the AI life.
19
When you're from another country and can't afford the free ChatGPT—I'm that poor
15
Hi everyone, I’m not familiar with neural networks – please scare me.
15
Помогите моему другу набрать людей в ДС сервер: discord.gg/6pEU9x8Xed

Присоединяйтесь к сообществу «GloWisP» на сервере Discord!

Global World Paradox | Участники: 13

discord.com
10
Load more