UT | $O

[UT | $O @utut0708]

OLLM V2 is here

Brings hundreds of zero data retention models and confidential AI with TEE
Introduces Dolphin Mistral 24B Venice Edition for a more open AI experience
Adds sandbox attestations for ORGN wửoktrees
@ollm_ai

[OLLM @ollm_ai]

OLLM V2 is finally here!

What's new:
∙ Hundreds of zero data retention models
∙ New TEE confidential models
∙ Dolphin Mistral 24B Venice Edition, for the most uncensored AI experience
∙ Sandbox attestations for ORGN worktrees

AI privacy, with proofs instead of hope.

[Origin @orgn_official]

If Mythos and Fable are sensitive enough to be export‑controlled and used by CISA to audit gov code, but a small team like ours can beat it on offensive cyber benchmarks in weeks at a fraction of the cost…

Maybe “frontier AI” is more narrative than scarcity.

[UT | $O @utut0708]

@orgn_official Truly amazing 😍

[Origin @orgn_official]

We built a cyber-capable AI model that, based on benchmarking and real-world testing, is perhaps the most capable offensive-security LLM we have seen

It scores 95% on Cybench, solving 38/40 tasks

But benchmarks only go so far, so we pointed it at real systems

Through authorized programs, we tested it against a Fortune-500 technology company, a major global telecom, and leading Web3 protocols.

The result: a 100% success rate - it found a viable path into every authorized target in our test set.

It adapts, reasons, and finds routes that aren’t obvious, even in systems that already passed audits and security reviews.

If this holds at scale, most current security assumptions break.

[Origin @orgn_official]

At ORGN, we know devs need tools that fit their workflow

That’s why ORGN is provider‑ and model‑agnostic

Bring your own key, plug in your own models of choice, and run them directly inside the ORGN IDE

[Origin @orgn_official]

We've reached a point where breaches can be fully agentic, but that doesn't make cybersec models & agents evil

The same capability that can crack your stack can defend it.

Don't let fear-mongering prevent you from using it to your advantage. Harden your own systems before attackers get to them.

That's what we're betting on. Stay tuned 🤝

[Origin @orgn_official]

A great take that also points to something partially related, but equally important:

Access to Intelligence should be universal

The same way access to knowledge should be
The same way access to education should be

A passport shouldn't decide whether you can access it.

[Origin @orgn_official]

Kimi k3 has arrived on ORGN.

Available now in the Gateway and CDE with full ZDR guarantees.

We’re already working to get K3 deployed on TEEs for hardware-enforced, attested confidentiality.

Stay tuned!

[UT | $O @utut0708]

@orgn_official Congrats . Big new 😍

[Origin @orgn_official]

ORGN is now a member of NVIDIA Inception

Our mission: to make agentic coding verifiably confidential. Total privacy for source code, credentials, and execution

A meaningful step toward making private AI practical at scale.

More to come!

#NVIDIAInception

[Origin @orgn_official]

Not all private AI is the same, so we make that distinction very clear.

When choosing your AI model in the IDE, you will see tags: ZDR or TEE

ZDR = Policy. The provider contractually commits to not retaining or train on your data.

TEE = Silicon. The model is deployed on Intel TDX & NVIDIA TEEs. Privacy is enforced by the hardware itself.

[Origin @orgn_official]

Another week, another AI data leak in the headlines.

The pattern is always the same: your prompts and data sitting in plaintext on someone else's server.

This is the exact failure mode ORGN closes. Here's how we do it 🧵

[Origin @orgn_official]

Most AI providers store every prompt you type

But that shouldn't matter when you can access 200+ models privately on ORGN Gateway with:

• Zero data retention
• Privacy filter
• Confidential Compute options

Available in both API and in-app

[Origin @orgn_official]

AI is being used to grow broccoli now

Props to OpenAI for this ad, though. It's creative, unexpected, but also a good reminder that AI models are finding their way into very real, very unexpected use cases.

Anyway, shoot us your wildest uses of ORGN below 👇

[UT | $O @utut0708]

@SuperMissO Wow miss O come back

[Origin @orgn_official]

Did you know you can verify ORGN’s confidentiality protections yourself?

Every sandbox and TEE model workload generates hardware-backed proof you can check publicly in ORGN Scan

To us, policies are just promises, but attestations are facts.

[UT | $O @utut0708]

@orgn_official Security should come first 😸

[Origin @orgn_official]

People love coding with powerful AI until they realise how good AI is at breaking into it too

Mythos/Fable are literally under export control for this

Maybe, the bottleneck isn't productivity anymore.
Maybe it's securing code.

[Origin @orgn_official]

Why would you let your coding agent store data when confidential worktrees exist?

• One isolated branch per task
• Scope exactly what the agent touches
• Session stays invisible, even to us
• Cryptographic proof it stayed sealed

Here's 30 seconds of it in action ↓

[UT | $O @utut0708]

Now you’ve got my attention.

Insane release + strong benchmarks usually means something big is around the corner.

Really curious to see what you’ve been cooking. 👀

[Origin @orgn_official]

We have an insane release coming

Benchmarks are much, much better than just "promising"

If you care about the frontier of AI/ML, make sure you're following

Stay tuned

[UT | $O @utut0708]

@orgn_official Haha me too