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: webai.com/waitlist

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: huggingface.co/spaces/webAI-…

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: webai.com/blog/inside-the-la…

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: huggingface.co/webAI-Officia…

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 . 3.6B parameters, running locally. Final: 114₇ = 60₁₀. Try it on @huggingface: huggingface.co/webAI-Officia…

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: webai.com/blog/meet-twil-lm3…

"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: webai.com/blog/webai-lands-u…

couldn't be us 👀 join the waitlist for access: webai.com/waitlist

Introducing webAI Frontline. Server-class AI. Running on a single iPad. Every technical industry runs on documents nobody can actually search. Manuals, procedures, regulations that go thousands of pages deep. Frontline turns them into an expert you can talk to. Ask a question and get the answer, cited to the exact page it came from. Fully offline, running on the device in your hand. AI that works where the work happens. Explore Frontline: webai.com/platform/frontline

TwIL-LM3 has been downloaded nearly 80,000 times since we released it. TwIL-LM3 judges reasoning. Give it a chain of logic and it can tell you whether the conclusion follows and the proof steps hold up, then explain in plain English. Specialist models like this usually run in the background. TwIL-LM3 can work directly with people while still handling the formal reasoning underneath. Here's where developers, researchers, and teams are using it: Judgment tasks — deciding whether a logical claim or proof step actually holds up rather than just sounding convincing. Mixed products — anywhere the same interface needs to answer everyday questions AND handle formal reasoning without switching models. Explanations — turning the output of a solver or formal reasoner into something a person can understand. TwIL-LM3 doesn't replace a real verifier for safety-critical work, but for everything else it gives you a checkable answer. It's fast, concise, and runs on your device. Get the model on @huggingface: huggingface.co/webAI-Officia…