Liquid AI reveals 8B-A1B MoE trained on 38T

226 points85 commentsa day ago
asb

Beware the license. They misleadingly state on the blog post "Open-weight — Download, fine-tune, and deploy without restrictions". But if you read their license <https://huggingface.co/LiquidAI/LFM2.5-8B-A1B/blob/main/LICE...> it has significant restrictions for any org with other $10M in revenue.

onlyrealcuzzo

I just tested this on a bug fixing benchmark I'm working on.

It did not perform as well as I expected. Qwen2.5-Coder-3B (2 years old) outperformed it by a wide range -> fixing ~50% of bugs whereas this model only fixed ~12%.

Granted, it's not a coder specific model, but given its benchmark performance to Gemma models, and that it's two years newer, and that it's an MoE with 8B total params, I expected it to be more competitive.

show comments
mlmonkey

Question: I have a dirty car and the car wash is just 50 meters away. Should I walk or drive to the carwash?

Answer: . . . . So, unless you have a compelling reason not to, walk to the car wash.

show comments
2001zhaozhao

At some point we have to be running into some inherent mathematical limits of knowledge compression, right? No way the knowledge benchmarks on these 8B models will keep getting better without overfitting on these benchmarks

show comments
SubiculumCode

Anybody use their localcowork [1] before? That is where the demo lives. Or not?

[1] https://github.com/Liquid4All/cookbook/tree/main/examples/lo...

adityashankar

This is super interesting, I'm particularly excited for this one as it may allow teams to scale this architecture for VLAs (vision language action models), and having sparser models means more real-time actions on a locally hosted model

demo link for anyone that wants to try this out https://playground.liquid.ai/chat?model=cmppnbgse000004l4bc8...

Ifkaluva

Liquid does amazing work, but I kinda feel like they are overtraining their models. 38T tokens seems like a lot for an 8B model

show comments
chabes

The small models are getting really impressive.

I recently realized that Qwen3.5:4B is way more capable than I thought a model that size could be.

Combine that with the work Liquid puts into RL and fine tuning, and you get models that perform extremely well on minimal hardware.

Combine that with your own fine tuning, and you get a specialized tool that is fast, private, and doesn’t require internet connection.

show comments
irthomasthomas

Woah, chinchilla scaling is 20 x active_params. I think mistral was 2 x Chinchilla. This is 1800 x

frankdlc222

Look at the accuracy numbers and these things clearly don't know much yet, and I'm not about to hand one my hardest work. But you can see where it's going. As quantization and the MoE stuff keeps getting better, "good enough to just run on my own machine" keeps eating into more of what I'm currently paying a frontier lab for. Once a local model can handle like 80% of what I need, the math stops making sense for the subscription.

kilroy123

Hmm, I asked it who made it, and it says Google?

show comments
grigio

I tested the previous model from Liquid, unfortunatly big claim but poor real performance

ramshanker

Guess we can run this even on CPU!

bee_rider

They seem… much better than all the models they compared against? What’s the catch?

show comments
elorant

Wow, this is fucking phenomenal. I fed it a long transcript asking it to create a summary and it executed it extremely well. For an 8B model this is quite impressive.

show comments
feelingsonice

Is Liquid AI still using the liquid neural network architecture?

HenryMulligan

Why does this not have (day-one) support for Ollama? The previous model is on there? Is it related to the ongoing refactor work or are people abandoning Ollama for other LLM engines?

show comments
gmuslera

Homeopathic AI

show comments
zmmmmm

No vision support?

jauntywundrkind

I really love how fast it is! Their press release comparing it on Strix Halo and M5 Max are impressive. It going twice as fast at GPU benchmarks even more so!