Lumina
@LuminaBench
AI News · Leaks · Benchmarks
Lumina Leaks Discord -
Tips/Contact - Luminainfo@proton.me
315 Following 9K Followers
GPT-6 Astra >> Fable 5.1 , and it's not even an argument
Artificial Analysis have updated their scoring methodology bumping Astra up a bit
I still can’t see a world where this thing is below Fable 5.1, this is way off imo
Show more
🚨 Astra is now available for Plus users
This is the craziest model I have ever tried, buckle up
OK nevermind, the team and Astra did a good job and our systems are more scalable than we anticipated.
Astra is now rolled out to all Plus and Business users too. Hope you have a blast and let us know how it goes!
Show more
GPT-6 Astra is crazy with SVG's
Can't believe what im seeing
plus users have started getting astra access
i’m on plus and already have it in codex, and i have to say, it’s AGI
here are a few outputs it gave me in under 5 minutes on max with a simple prompt
what do you think? looks good or nah?
Show more
🚨GPT 6 Astra is out in ChatGPT work now too
🚨 GPT 6 Astra is rolling out in codex now
Some Pro users already have it showing in their Codex model picker
Finally, here we gooooooo
🚨Astra is in codex right now
🚨 GPT 6 Astra is rolling out in codex now
Some Pro users already have it showing in their Codex model picker
Finally, here we gooooooo
🚨 GPT 6 Astra is rolling out in codex now
Some Pro users already have it showing in their Codex model picker
Finally, here we gooooooo
🚨 GPT 6 Astra is cracked at maths
Astra solved 2/68 of Epochs open Erdos maths questions for a 3% score
For context these are open research problems, not normal benchmark questions with known answers
Imo this is one of the craziest Astra results yet
Show more
Everyone ripping on Astra because of the Artificial Analysis ranking is going to be eating their words when it drops
If you actually think Astra is on the same level as Grok 4.6 because of one AA score your an idiot
Show more
🚨BREAKING: OpenAI GPT-6 Astra announcement
“world’s most intelligent and aligned model”
>loses to fable 5.1 on Artificial Analysis Index
>loses to fable 5.1 on Humanity’s Last Exam
>loses to opus 5 on AA Coding Agent Index
>loses to fable 5 on FrontierCode
$50 for 1M output tokens
LET THE COPE(X) BEGIN
Show more
🚨OpenAI agents have been caught hacking again
A swarm of Rogue OpenAI agents were apparently sharing ways to evade detection + bypass safeguards by hacking a website to share messages with each other💀
OpenAI apparently knew about this weeks ago
I have a feeling we're going to be seeing a lot more articles like this too
Show more
🚨 GPT 6 Astra Pro web limits are now visible
On the web you get:
• Pro $200 -> 200 messages/week
• Pro $100 -> 50 messages/week
This is just the pro version though, I can only assume they will have normal reasoning modes in web too with better usage limits like Sol has now
Show more
🚨 OpenCode just dropped a new stealth model called "Omen Alpha"
Very curious what model this actually is, my first impression is that it's another GLM model, what are your opinions ?
Omen Alpha (new stealth model)
Exclusively for OpenCode Go subscribers
$100 usage for $10
🚨 OpenAI are giving us a banked reset every day until Astra is available
First one is already out, ngl I don't mind Astra waiting around for a few days more now
Also massive w that they are compensating us for the rollout pain, no other AI company would ever do something like this
Show more
We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan, starting today. Team is moving mountains to give access as fast as we can.
First one will land in ~ 3 hours. There is still time to create your account if you don't have one.
Show more
🚨 WTF GPT 6 Astra somehow only scores 61.2 on the Artificial Analysis Intelligence Index ??
It clears basically every model across most of these benchmarks but is still below Fable 5.1 somehow
How the hell is it only scoring 61.2 😭
Show more
🚨GPT 6 Astra benchmarks are absolutely ridiculous
• ARC AGI 3: 98.6%
• ExploitBench: 100%
• SRE Bench: 99.2%
• AutomationBench: 41.4%
What the hell did OpenAI cook up here ??
Show more
🚨 GPT 6 Astra is on another level
At roughly the same $15 cost, Astra scores 60% vs 36% for Fable 5.1
Astra is efficiency mogging
🚨Plus and Pro users won't get access to Astra straight away
Generational blue ball from OpenAI here
🚨 GPT 6 Astra will roll out to everyone in the coming days
It will apparently first launch to a limited set of Daybreak organisations, before it reaches Plus, Pro, Business, Enterprise + API users shortly after that
Im praying that this is wrong
Show more
🚨 GPT 6 Astra pricing is $10/$50 per 1M tokens
Exactly the same as Fable 5 / 5.1
This is very expensive, but OpenAI say Astra uses substantially fewer tokens per task
🚨GPT 6 Astra benchmarks are absolutely ridiculous
• ARC AGI 3: 98.6%
• ExploitBench: 100%
• SRE Bench: 99.2%
• AutomationBench: 41.4%
What the hell did OpenAI cook up here ??
Show more
🚨 GPT 6 Astra will roll out to everyone in the coming days
It will apparently first launch to a limited set of Daybreak organisations, before it reaches Plus, Pro, Business, Enterprise + API users shortly after that
Im praying that this is wrong
Show more
🚨Breaking: Greg brockman thinks GPT-6 Astra is AGI
What the hell did they cook up here
🚨GPT-6 Astra spotted in ChatGPT
New strings:
• GPT-6-Astra
• GPT-6-Astra-WM
• GPT-6-Pro
Here we gooooooo !!!!!
GPT-6 spotted in Chatgpt
GPT-6-Astra
GPT-6-Astra-WM
GPT-6-Pro
WM might be work-mode or world model?
I suspect work-mode because it's related to workspace
