Your AI creative suite. Every Frontier model. Dozens of proprietary models. E2EE privacy. Web, Discord, Studio & Katana API.
Joined December 2022
- Tweets10.6K
- Following2.4K
- Followers25.8K
- Likes24.9K
We have Nano Banana 2.1+ at home now
4K image generation in 10 seconds! ⚡
Meet MiniMax H3-Image.
16 million pixel resolution. Stunning realism. Precise prompt following & editing. Up to 16 reference images. All in 10 seconds.
And it's open-weight. 🚀
Available now via our API, MiniMax Code, and Hailuo AI — with 50% off 16MP+ generations during launch week!
Can't wait to see what you create. 🔥
Hugging Face: huggingface.co/MiniMaxAI/Min…
API: platform.minimax.io/docs/gui…
MiniMax Code: code.minimax.io/about
Something a bit different, to hopefully save us and someone else from the depths of GPU eBay
We've recently swapped out a batch of RTX 4090 GPUs (Windforce, dualslot blower shroud fitted) in our hardware stack.
All GPUs are in full working order, and tested/verified by TechBuyer with documentation enclosed.
Available as individual units, or for bulk purchase. Happy to come in at 10% under the eBay market pricing, and willing to talk further discounts with anyone looking to pick up a few!
Drop into our DMs or fire us off an email on info[at]imgnai[dot]com if you're in the market!
absolute banger from the sad bois 🫡
proud to be your official shitposting partner of choice
every gen in this video was made on @imgn_ai
every frontier model + their own models, a free video editor in your browser, and it's cheaper than Venice and Higgsfield
official AI partner of emonad:native for a reason
Emonad
Work is still ongoing on our HyperSnap (real Farcaster) client, hoping to have a release out in open beta very soon!
As noted, red-teaming/security is a pretty heavy focus for us on building, which unfortunately can slow things down quite a bit.
LLMs fine-tuned for cyber are getting (/already are) extremely capable, even moreso in a pair of experienced hands. As such, red-teaming has become a much higher focus for our development cycles in recent months.
Feature-wise, focus remains on parity with Farcaster + HyperSnap native protocol features.
There'll be room to add lots in the future, but we're looking to keep this particular client aligned with HyperSnap native or governance-driven features, with our main focus outside of that being on stability/speed/security.
And we're back - with apologies for the slightly longer than expected maintenance period!
We're back up and running, and should be feeling even faster than ever - along with having a bunch of ux improvements to boot.
Note that while you should hopefully just be feeling a faster and more responsive imgnAI, a huge amount of our core infrastructure was updated during this maintenance window.
As such, we expect there may be some (temporary!) teething issues over the next few days while we finish getting all the kinks ironed out.
Once again, we appreciate everyone's patience as these big updates roll out - especially when the maintenance window can be a little more generous than intended 🫣
All for good reason of course - more news coming soon! 🫶
Apologies for the delays - working hard on bringing everything back online!
Along with general maintenance & hardening (in this the era of daily CVEs), we've also some additional ux improvements & speedups scheduled to go live - so, everything should be that little bit better once we're back up and running.
Shouldn't be too much longer left - we'll alert you on here once we're back live and kicking!
endgame btw
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
We’re living in the age of autobuild abundance
If you’re in a lowcap meme and they’re not building out stuff like this, they’re just too lazy to succeed tbh
If you’re not in a lowcap meme and need one on your balance sheet, look no further than the ones working both for their bags, and by association, imgn bags
emo bot made a brainrot studio to use with @imgn_ai while i was making a pot roast.
brother is cooking rn.
gud bot tbh.
emonad
You already knew @higgsfield were scammers, but this really drives home just how bad they are.
Funniest bit: their AI Movie which debuted at "Cannes" - it didn't debut at the film festival. They debuted it in Cannes (the City).
The next gap we'll see will be engineers connected to their coding agents via telepathy.
And if you don't have it, you're going to be like someone coding by hand vs. someone coding with Astra or Mythos-level LLMs today.
Neuralink’s first product is called Telepathy.
It enables high-speed communication with computers or humans simply by thinking. Output bandwidth will vastly exceed the fastest human speech or typing in a few years. Read that sentence again.
2
15
1,003
Your daily reminder that @higgsfield is a scam platform which relies on paying influencers to break @X's terms of service in posting undisclosed promotional content.
Higgsfield only serve third-party models, at an enormous price premium.
Shop elsewhere, or build local.
Last night, @higgsfield sent @ComfyUI a cease and desist (attached)
They want the Open Source Higgsfield challenge taken down within 48 hours.
Their argument is that our use of the Higgsfield name and that using their CEO's likeness is an infringement violation (man, the irony). Claiming that it makes the event look like Higgsfield is associated with it.
This couldn't be farther from the truth, which is that Higgsfield made a promise to open source its product and never did.
For the past two weeks, builders have been recreating popular closed apps as open workflows anyone can run.
So we asked our community to deliver where Higgsfield failed: apps in the open, built by the community, for the community. We've made it clear that the challenge is organized by ComfyUI and not affiliated with or endorsed by Higgsfield.
When we started Comfy, we made the community a promise. Our product stays open, free, and yours. We’ve kept it every day since, and one letter isn’t going to change that.
The challenge stays up. Comfy and open source will win!
