What Language Model do you need?

It’s been 2 months since I started working at Liquid AI (aka Liquid).

In case you just landed on Planet Earth after a 15-year trip around the Cosmos and somehow missed the memo: Liquid is one of the top AI labs in the world, famous for its hardware-aware Language Models (the LFMs) that run like a bullet.

And yes, it’s full of ridiculously smart people doing all sorts of science, engineering, and what I can only describe as computational alchemy:

• Pre-training models from scratch

• Mid-training models (yes, that’s actually a thing!)

• Post-training models using SFT, DPO, GRPO and a long list of exotic 3-4 letter combinations.

• Generating synthetic data like caffeinated data scientists who forgot sleep exists

• Creating no-code tools that make model customization feel less like dark magic

• Building Android and iOS SDKs to make deploying small language models as easy as calling an API

• Optimizing inference engines to squeeze out every precious millisecond of latency and megabyte of RAM

But you know what? That’s not enough.

All this grey matter and computational firepower is useless if we’re not building the things YOU actually need.

So here’s my question for YOU:

👉 What Language Model do you want Liquid to release next?

Be as specific as you can, for example

• A Vision model for medical imaging with less than 5B parameters?

• 4-bit quantized model for Raspberry Pi running object detection under 2GB RAM?

• Multilingual model for legal document Q&A with <3B parameters supporting EU languages?

• Code model under 7B parameters that understands legacy Python 2.7 codebases?

• Tiny model for ARM Cortex-M microcontrollers that fits in 100KB flash memory?

• Reasoning model with <10B parameters that solves competition-level math problems?

• Multi-modal model under 8B parameters processing PDFs, images, text, and audio simultaneously?

• Agent orchestrator like NVIDIA’s 8B but under 3B parameters with <100ms routing latency?

Drop your answer in the comments below.

And if you are too shy to share a public comment here, join the Liquid AI Discord Community and send me a private Direct Message there.

Don’t be shy.

I’m all ears (and I promise to read every single response).

Pau