webAI

Sovereign AI for enterprises. Build AI you own on infrastructure you control. Private, specialized models that run where your data lives.

[webAI @thewebAI]

LLMs are getting more rambly. We built TwIL-LM3-Pro to go the other way.

Less than half the tokens of VibeThinker-3B and Qwen3-8B. 1.7x and 5.4x faster answers on general reasoning benchmarks.

Full benchmarks: https://t.co/dr1LDchxMr

[webAI @thewebAI]

AI shouldn’t think for you. It should help you think for yourself.

The things that make your work different, the things that make it worth paying attention to, aren’t on the internet. They live in you.

That’s what webAI is built on: https://t.co/K1nMfI41W7

[webAI @thewebAI]

Most of the AI conversation is about scale. Bigger models, more parameters, longer context windows. The assumption being that bigger is smarter.

We’ve been building toward the opposite.

TwIL-LM3-Pro is a 3.6B reasoning model that handles formal problems models 7 to 10x its size struggle with, and it runs on a laptop.

Small isn’t a compromise. For the problems we care about, it’s the point.

A model that fits in RAM, answers in seconds, and can be understood end to end gets us closer to what AI should actually be: specialized, local, and yours.

Download the model on @huggingface: https://t.co/bZ8ND3PMKP

[webAI @thewebAI]

LLMs are getting more rambly. We built TwIL-LM3-Pro to go the other way.

Less than half the tokens of VibeThinker-3B and Qwen3-8B. 1.7x and 5.4x faster answers on general reasoning benchmarks.

Full benchmarks: https://t.co/dr1LDchxMr

[webAI @thewebAI]

How does a 3B parameter model compete with frontier models?

In our latest video, our Intelligence Lab team walks through how they built the TwIL reasoning models. The thesis behind formal reasoning, the moment that reshaped their data pipeline, and where the lab is headed next.

Watch the full video: https://t.co/A7Cahkwx1J

[webAI @thewebAI]

@Webmxzora 👀

[webAI @thewebAI]

We ran TwIL on a CPU with 15 GiB of RAM. No GPU. It generated tokens 98x faster than Qwen3.8-27B on the same hardware.

The test:
- Intel Xeon CPU, 4 threads
- 15 GiB RAM
- No GPU

Results:
- TwIL-LM3-Pro (2.24 GB weights): 10.47 tokens per second
- Qwen3.5-9B (5.67 GB weights): 4.36 tokens per second
- Qwen3.8-27B (16.45 GB weights): 0.107 tokens per second

TwIL and Qwen 9B both run at the RAM's memory limit of about 24 GB/s. Qwen 9B is 2.4x slower only because its weights are 2.5x larger.

Qwen 27B's 16.45 GB of weights don't fit in the 15 GiB of RAM. Parts have to be re-read from disk for every token. Size alone would make it 7.3x slower. The disk reads add another 13x. Total: 98x slower than TwIL.

With enough RAM, the gap would shrink to about 7x. But most people don't have unlimited RAM, and most people don't have a top-tier GPU. The model you can actually fit matters more than the one you can't.

Download TwIL-LM3-Pro on @huggingface: https://t.co/pTRuIDyV0L

[webAI @thewebAI]

LLMs are getting more rambly. We built TwIL-LM3-Pro to go the other way.

Less than half the tokens of VibeThinker-3B and Qwen3-8B. 1.7x and 5.4x faster answers on general reasoning benchmarks.

Full benchmarks: https://t.co/dr1LDchxMr

[Devin White @DevinWhiteAI]

We’ve been playing around with the new TwIL-LM3-Pro model and in this maze problem it solved it so fast and even beat Qwen3.8-27B!

Seeing this model perform this good at only ~4B is crazy, definitely should check it out if you haven’t already!

[webAI @thewebAI]

Base-7 arithmetic is a brutal test for small models.

There's no base-10 shortcut, so every carry has to be reasoned through directly in the target base. Most 3B models either skip the carry logic or hallucinate the conversion back to base-10 at the end.

TwIL-LM3-Pro self-corrects a wrong carry mid-trace inside <think>. 3.6B parameters, running locally.

Final: 114₇ = 60₁₀.

Try it on @huggingface: https://t.co/JyiFTWJySy

[webAI @thewebAI]

LLMs are getting more rambly. We built TwIL-LM3-Pro to go the other way.

Less than half the tokens of VibeThinker-3B and Qwen3-8B. 1.7x and 5.4x faster answers on general reasoning benchmarks.

Full benchmarks: https://t.co/dr1LDchxMr

[David Stout @Davidstout]

AI models are not conscious. A CEO or leader who seriously claims that current systems are conscious is either profoundly confused about the technology or engaged in deliberate misrepresentation, and is unfit for authority over AI development or deployment.

Current models are large statistical systems trained to predict tokens. They have no evidence of subjective experience, unified agency, persistent self-models independent of the prompt, or any of the functional markers that would be required even on generous functionalist accounts of consciousness. Anthropomorphic outputs are produced because the training data is full of human self-description; that is imitation, not evidence of an inner life.

Treating such claims as a reason for removal from power is therefore not a medical diagnosis. It is a judgment about competence and incentives.

Leaders who assert consciousness in present systems either do not understand what they are shipping, or they benefit from the confusion—through valuation narratives, regulatory positioning, or public relations. In either case they are poor stewards of a technology whose real risks (capability, misuse, concentration of power) are already concrete and do not require speculative inner experience to matter.

[webAI @thewebAI]

Download TwIL-LM3-Pro on @huggingface: https://t.co/m36fNzYKpw

[webAI @thewebAI]

LLMs are getting more rambly. We built TwIL-LM3-Pro to go the other way.

Less than half the tokens of VibeThinker-3B and Qwen3-8B. 1.7x and 5.4x faster answers on general reasoning benchmarks.

Full benchmarks: https://t.co/dr1LDchxMr

[David Stout @Davidstout]

Half a million downloads in a month. Today, our open source family takes another step forward.

Thank you for the incredible support behind our first-generation models. We’re excited to introduce TwIL-LM3-Pro.

At just 3.6 billion parameters, it brings powerful reasoning to everyday computers, with quantized builds that run locally. No cloud required.

In our evaluation:
Formal logic: Highest recorded headline score among the small models compared—beating China’s VibeThinker-3B by 35% and Qwen3.5-4B by 24%, and Liquid AI’s LFM2.5-8B-A1B by 47%.

Broader reasoning: 95% on SVAMP and 64.1% on MuSR, the highest recorded scores among the small models compared.

BIG-Bench Hard’s logic subset: 95.4%, compared with VibeThinker-3B’s 61.1%.

We believe AI is entering a post-training era. The advantage will increasingly belong to companies with the best pipelines and those that can produce capable, personalized intelligence faster and more efficiently, then put it on devices people already own.

That’s what we’re building at webAI. And we’re only beginning to share what’s coming out of our lab.

Coming soon: Meridian, our family of frontier-class models built to run on device. Our most advanced models will be available through the @thewebAI application.

Join the waitlist as we expand access. Proudly built in Austin, Texas. 🇺🇸

[Morning Call @CNBCMorningCall]

Jason Rathje, President of Public Sector at @thewebAI, discusses U.S. China AI competition, open source models, national security and the need to build resilient American AI supply chains.

$NVDA $AMD $AVGO $INTC $TSM
@MorganLBrennan

https://t.co/4Xe7pIbHWs

[webAI @thewebAI]

"A general model is only the beginning. Individuals and organizations need intelligence deeply specialized to them." — @Davidstout

Today we're partnering with Forge to bring private, specialized AI to enterprise customers.

Read the announcement: https://t.co/LF6VmYEua5

[webAI @thewebAI]

"More AI is not synonymous with more data centers."

@Davidstout, our CEO, in a new op-ed for the @dcexaminer on why the trillion-dollar AI data center build-out is a bet on the wrong future.

The future of AI is smaller, specialized, and local.

Read the full piece: https://t.co/CZ8cxlpnpN

[webAI @thewebAI]

"Many open models can now be powered by a phone or laptop. That has made the technology more accessible and even cheaper. Leading closed models are getting more complicated and requiring more computing power."

@Davidstout, our CEO, in a new @nytimes piece by @elitanjourno.

Open, small, and specialized is where AI is going.

Read the full piece: https://t.co/NQZNI4cl1W

[David Stout @Davidstout]

I think the answer is not to slow down. It’s to invest in approaches our adversaries cannot easily replicate and cannot fully embrace without changing how they exercise power.

The frontier labs have built a paradigm around increasingly powerful centralized models and enormous concentrations of compute. My concern is that we’ve defined the competition around a playbook our adversaries can follow, rather than an advantage they will struggle to reproduce.

At @thewebAI we’ve been working toward a civilization of intelligence: distributed, specialized intelligence grounded in the specific, tacit knowledge of people, teams, and institutions. Our Intelligence Delivery Network is part of what connects that intelligence without requiring all knowledge and control to be centralized.

The distinction is not just distributing compute. It’s distributing the ability to understand, decide, and act.

Our strength is trusting people throughout the chain of command to exercise judgment not simply follow instructions. That principle of trust and decentralized execution is already foundational to our military doctrine. We should build intelligence that amplifies it.

Our bet is that this creates a fundamental dilemma for adversaries whose power depends on centralized control. They can pursue the same technology. But realizing its full value means empowering people and institutions to act without every decision flowing through the center.

That is much harder to replicate than a model. Copying the technology does not recreate the accumulated knowledge, relationships, and trust that make the system valuable.

We should not slow down to preserve a lead in a paradigm others can copy. We should invest in one that turns our greatest strengths into an enduring advantage.

[webAI @thewebAI]

On the 25th anniversary of September 11, we remember the nearly 3,000 lives lost, the families who still carry that day with them, and the heroes who ran toward danger to help others.

We're grateful to all those who have defended our country and freedoms, then and now.

We will always remember.

[webAI @thewebAI]

couldn't be us 👀

join the waitlist for access: https://t.co/K1nMfI41W7

[David Stout @Davidstout]

@thewebAI is up.

[David Stout @Davidstout]

The most valuable AI company won’t be built around a single model. It will be built around the system that can create the next thousand.

The physical world won’t run on one giant model in the cloud. It will run on small, specialized models deployed everywhere.

At @thewebAI, that system is now producing N expert reasoning models every week and accelerating.

Meridian is next.