Astro Polo
@astropol0
AI Tracker • Part-time Youtuber | AI breakthroughs, real money making opportunities & Big Tech critique | | DMs open for collabs & Deals 📥
1.5K Following 3.3K Followers
Okay now things getting really out of hand
They might really skip Grok 4.7…
Introducing Grok Voice Transcribe 2.0. It’s the world’s most accurate speech transcription model.
If they can’t deliver Grok 4.7... give Grok users a reset at least
super frustrating for Grok users...
Don’t know if this sheet is genuine...
somehow it has Gemini 4 Pro beating Astra, Fable 5.1, Opus 5 and Sol on almost every row… at $2.25 / $11.25
i mean if even half of it is real... Google just walked back into the fight and if it’s slop,, it’s very good slop
october is going to be chaos fr
Show more
Saw pretty crazy Gemini 4 Pro results yesterday....
surprised that nobody from DeepMind is posting anything like they used to vaguepost every checkpoint, now the Arena clips are doing laps and the official accounts are dead silent
this can mean two things.. either the model is usually close… or the last one flopped and they learned not to vagueposting
Show more
Grok 4.7 is in testing and the first looks aren’t promising..
reportedly another cheaper version from SpaceXAI with a little improvement over Grok 4.6
delaying it might be the best move... people won’t take another same-like model
they really should skip it and focus on 4.8 so the upgrade actually feels good...
Show more
SpaceXAI should skip Grok 4.7 and put everything on 4.8...
Since Grok 4.7 is a patch on the old RL that's why it quits early...that’s why multimodal is still broken.. that’s why they keep delaying it...
while 4.8 is the actual upgrade with 2.5T parameters plus new C++ stack from the Starlink team...
shipping a mid model just to hit a tweeted date wastes the week
that's why they should skip the patch and ship the stack instead..
Show more
Okay Grok bot now can talk
voice is rolling out on desktop and mobile over the next couple of days but....
where tf Grok 4.7 is @SpaceXAI
qwen3.8-omni-flash is now out...
Alibaba’s cheap & fast omni model.. one brain that takes text + image + audio + video and answers in text,, about 25% up vs the last Omni
they also map it as the Gemini 3.8 Flash class swap for audio/video understanding
Show more
Google is soo back....
Gemini 4 Pro checkpoint is lighting up stronger outputs people are calling it the thing hiding behind a Flash label....
PS5 SVG, Three.js rockets, Full interceptor scenes on a first pass...
testers sayin it looks more capable than astra and fable 5.1
this might actually be the comeback people were waiting for just showed up on the arena
Show more
Google’s redemption arc is here?
can’t believe Gemini 4 Pro is giving this much output in an Arena checkpoint…
Full interactive SVG
Pelican on a bicycle
Inverse kinematics
Scene switching (sunset → day → starlight).
Bell, pedals, vector-perfect
took ~6 minutes.…
Show more
🚨 Gemini 4 Pro (Pelican riding Bicycle SVG)
This is very impressive and only took around 6 minutes to make in total
The new Gemini 4 Pro checkpoint is really impressive
Looks like Grok 4.7 is coming….
base_model:grok-4.7 just showed up in Google Cloud quotas
if you don’t know this is the same pattern as previous Grok drops on Vertex.. quota lands first, model follows
after all the delays we might actually get it today hope it’s worth the wait and a usage reset might come with it so get ready to test…
Show more
no GPT-6 Sol this week
the ship left the harbor, circled twice, then quietly docked again.. lol
Worth the wait only if we get one weekly resets
the main thing i was excited about launching this week will be next week instead, but imo worth the wait!
🚨 Gemini 4 Pro checkpoint under testing on
SpaceXAI should skip Grok 4.7 and put everything on 4.8...
Since Grok 4.7 is a patch on the old RL that's why it quits early...that’s why multimodal is still broken.. that’s why they keep delaying it...
while 4.8 is the actual upgrade with 2.5T parameters plus new C++ stack from the Starlink team...
shipping a mid model just to hit a tweeted date wastes the week
that's why they should skip the patch and ship the stack instead..
Show more
First European frontier?? 💀
It’s time. Live on OpenRouter. Zero data retention. Frontier for free.
Union Alpha is getting mixed feedback
the chart looks like Astra but some testers say the product does not...
good side: DeepSWE around 74. Sits near GPT-6 Astra high on Terminal-Bench.. free week.. No training on your data and works in every harness....
bad side: slow as hell for a lot of people, 7.6 tok/s... 78 seconds for 67 tokens An hour for 13 tool calls.. hides its work, one guy couldn’t even finish an SVG...
same Ox Alpha playbook... green box first queue second
Show more
A nameless stealth model just sat down next to Astra...
Union Alpha live on OpenCode free for the next week built for agentic coding... Images work and no training on your data
It currently sits next to GPT-6 Astra (high) on Terminal-Bench at anticipated pricing that would put it in the cheap and good quadrant
same playbook as Ox Alpha....
Show more
Elon should compensate us for the Grok 4.7 delay with a usage reset
Grok 4.7 comes out in 10 days
A nameless stealth model just sat down next to Astra...
Union Alpha live on OpenCode free for the next week built for agentic coding... Images work and no training on your data
It currently sits next to GPT-6 Astra (high) on Terminal-Bench at anticipated pricing that would put it in the cheap and good quadrant
same playbook as Ox Alpha....
Show more
Union Alpha (stealth model) is free for the next week
- no data training
- built for agentic coding
- supports images
let's see what you can do
Google DeepMind just published Dream-RSI....
it recursive self improvement through evolving worlds
not the model rewriting itself into God but the explorer learning how to search
coding agents waste a fortune trying kernels, algorithms, math... fixed search goes stale and tuning search online means paying the agent again
Dream-RSI treats the discovery tree the agent already built as a world replay that world... test thousands of new search policies with zero new runs
deploy the winner.... It finds new stuff that tree joins the pool and it Loops
Show more
Holy shiii.. Apple’s new Siri AI just passed the strawberry test!
If you don’t know ChatGPT spent years failing this 😂
we are probably getting two cheap and efficient models this week
Grok 4.7 and GPT-6 Sol
