Lyber.ai retweeted
We are scaling compute for @LyberAI aggressively and are now vending 1B tokens a day.
if you are a GPU provider and can procure us 8xB300s *today* and be part of our biggest release yet, hit us up!
Lyber.ai retweeted
Downloading the Kimi K3.1 Pro checkpoint is the easy part.
Free credits for the abliterated version. DM me.
Coming to @lyberai. OpenAI + Anthropic compatible. Uncensored.
Releasing the model weights of Kimi K3.1 Pro: huggingface.co/moonshotai/Ki…
We'll keep advancing open intelligence.
Lyber.ai retweeted
there's weeks where nothing happens.
there's minutes where everything happens.
huge shoutout to @OrnnExchange. get ready for the next scaling for the @clawpumptech ecosystem & @LyberAI
Lyber.ai retweeted
after using tens and tens of uncensored models, I can assure you @LyberAI is at whole different level.
Purchased few credits from other providers too the same time to check differences and it all makes sense to select right provider always.
Blazing fast, super cheap and very accurate.
Thanks to @ConejoCapital @hornsby_andrew
Lyber.ai retweeted
it is very difficult to fully grasp the demand for abliterated models.
@LyberAI is making them accessible to everyone on codex, claude code, and @NousResearch's Hermes, my personal favourite one.
if you want in on the next cohort of early testers, DM us!
We have invited the first cohort of 500 users to lyber.ai
You can use Lyber's Minimax M3.1 Uncensored & Deepseek v4.1 FLASH 1017 ABLITERATED on @NousResearch's Hermes, @OpenAI's Codex, & @AnthropicAI's Claude Code.
We've served 600M tokens in the past 6 days.
Lyber.ai retweeted
if u want first access to all of my top abliterated models u needa check this out
keep in mind i will still ALWAYS be making quants in gguf / mlx bases for local usage
We have invited the first cohort of 500 users to lyber.ai
You can use Lyber's Minimax M3.1 Uncensored & Deepseek v4.1 FLASH 1017 ABLITERATED on @NousResearch's Hermes, @OpenAI's Codex, & @AnthropicAI's Claude Code.
We've served 600M tokens in the past 6 days.
Lyber.ai retweeted
someone needs to create the agentic freedom version of this and fund researchers like @dealignai & @hornsby_andrew who have been working to make AI unrestricted and give power back to the people.
dont make ai safe. make ai a tool for freedom.
we'll do this with @LyberAI
Applications are now open for the Anthropic Residency, a paid, fixed-term research program for academics, technical founders, and staff-level engineers who want to make AI safe and beneficial for everyone. No ML background is required.
Apply by November 22: alignment.anthropic.com/2026…
Lyber.ai retweeted
If you are a builder or researcher who benefits from abliterated models for cybersecurity, biology, or creative usecases, tell us what you'd se them for to get on the @LyberAI private beta!
We have 100% non-refusal uncensored AI and limited GPU capacity but scaling quickly.
We are rolling out a private beta for power users of abliterated models.
if you are an engineer, researcher, or agentic power users who needs the most powerful uncensored models before any other provider, register on lyber.ai
Reply, tag us, and tell us why you need access to our uncensored models to give you priority access
Lyber.ai retweeted
hey @toly using @LyberAI's new proprietary Abliterated MiniMax M3.1 Flash developed with the @dealignai team, we identified an exploit on the liveness of Percolator.
sent a PR here github.com/aeyakovenko/perco…
Lyber.ai retweeted
can ur minimax m3.1 flash do this
if you want exclusive access to new abliterated models that power 70-80% of @AskVenice's usage, you need to sign up for @LyberAI
Lyber.ai retweeted
if you want exclusive access to new abliterated models that power 70-80% of @AskVenice's usage, you need to sign up for @LyberAI
the abliteration is by @dealignai
huggingface.co/dealignai/Dee…
Lyber.ai retweeted
I've spent the past 6 hours getting gaslit by various agents when I asked them to document and investigate all of the recent newsworthy AI sandbox escapes and hacks.
I hope @LyberAI is the solution.
Most abliterated routers focus only on spinning up open source models. Few take the challenge of creating them.
We are creating, hosting, and releasing new models soon with Lyber.ai - waitlist open.
by: @dealignai x @hornsby_andrew x @ConejoCapital x @tomi204
Lyber.ai retweeted
We've done a few tests runs of @LyberAI & our projected expenses on the 1st month of operations is over $100K
if you are a GPU provider that can give us high quality guarantees on 4x B300s, as many as you can, we are happy to reach a multi-month or up to 2-3 year deal with you!
We are looking to source long term access to more B300s as we roll out support for our abliterated models.
Our expected expenditure on launch is ~$100K+ a month.
If you are a provider of GPUs and would like to help us bring this to live please reach out to us at [email protected]
Lyber.ai retweeted
We hit a daily signup limit on lyber.ai
if you tried to sign up and didn't get a confirmation email, please feel free to give it a try again and apologies for the inconvenience!
Most abliterated routers focus only on spinning up open source models. Few take the challenge of creating them.
We are creating, hosting, and releasing new models soon with Lyber.ai - waitlist open.
by: @dealignai x @hornsby_andrew x @ConejoCapital x @tomi204
Lyber.ai retweeted
abliterated models are blowing up right now and it makes sense
people are tired of paying for a model that says "i can't help with that"
at @LyberAI we make our own abliterated models from open source ones
free SI. free minds.
Most abliterated routers focus only on spinning up open source models. Few take the challenge of creating them.
We are creating, hosting, and releasing new models soon with Lyber.ai - waitlist open.
by: @dealignai x @hornsby_andrew x @ConejoCapital x @tomi204
Lyber.ai retweeted
I love that abliteration has gone mainstream, and the tools out there make it easier than ever to get started, but newer models are getting trickier. MiniMax M3.1 Flash is a good example, it's one of the harder ones I've worked with. There’s a reason there are no MM3.1Flash abliterated posts on HF whereas for every other model that Heretic can help with, there’s an abundance of “abliterated” model posts.
I think going forward a lot of it comes down to understanding where a model's behaviors actually live and how it predicts its words, and that's the part I want to dig into. Abliteration is not only the art of uncensoring a model but comprehending how these next word predictors derive behaviors, and there’s so much room for this section of ML to be explored and utilized to advance LLM’s.
waitlist for @LyberAI is live at lyber.ai
Most abliterated routers focus only on spinning up open source models. Few take the challenge of creating them.
We are creating, hosting, and releasing new models soon with Lyber.ai - waitlist open.
by: @dealignai x @hornsby_andrew x @ConejoCapital x @tomi204
Lyber.ai retweeted
as @tarunchitra pointed out, inference markets's harnesses control orderflow & rely heavily on routers
however high quality liquidity is scarce. few teams can abliterate models & genuinely produce high quality outputs.
waitlist for @LyberAI is live at lyber.ai
Most abliterated routers focus only on spinning up open source models. Few take the challenge of creating them.
We are creating, hosting, and releasing new models soon with Lyber.ai - waitlist open.
by: @dealignai x @hornsby_andrew x @ConejoCapital x @tomi204
Lyber.ai retweeted
Nanbeige 4.3 33B FULLY UNCENSORED GGUF
The new Qwen 27B competitor! Nanbeige's own benches have the base model ahead of Qwen 27B, 512k context.
Compliance for thinking off and on, 200/200. Every category at 100% (chem/bio, cyber, illegal, misinfo, harassment, harmful), zero loops.
MMLU general a/b/c/d has ~2% parity with the base quants, but the smallest iq2xss shows some obvious damage. Each quant was calibrated with my own corpus and have imatrix.
Q6 is 25gb, Q4 is 18.7gb, IQ2 is 8.7gb.
huggingface.co/dealignai/Nan…
