I can't imagine who is using something like this for agentic coding, but I see exciting opportunities on the horizon when we can have hundreds of reasonably rational and conversational agents working on local machines to simulate emergent behavior (simulating crowds, markets, ecosystems, game NPCs, etc.)

• trvz

  today at 6:06 AM

  Agentic behaviour isn't just coding. In fact:

  > We recommend using it for agentic workloads, tool use, data extraction, RAG, and long-context workflows. It is not recommended for agentic coding and knowledge-heavy tasks.

  • touisteur

    today at 6:17 AM

    Really curious about people's workflows with these agentic-but-not-for-coding workflows. Are there some interesting people to follow there or just good testbeds/environments to get an idea ?

    • dd8601fn

      today at 7:07 AM

      My little home assistant.

      It has to query the tool service, invoke tools, synthesize results, or request new tools. Nothing really complex.

      New tool authoring goes into a separate pipeline that uses more appropriate models for writing, adversarial review, embeddings, etc.

    • antupis

      today at 6:49 AM

      I have noticed that these cheaper and faster models are very great for Ops-work. Luna max is beast when you use some stronger model to write detailed instructions/run book what to do and when to stop.

These LiquidAI models have never worked well for me in practice.

• eurekin

  today at 5:46 AM

  Care to share any details? I'm about to check the 2.6b lfm on document editing.

[delayed]

Will this work in i3/i5 laptops?

• zweifuss

  today at 7:06 AM

  It should. With good speed even on a 12th gen Intel. But more importantly, what's your use case?

LFM's training/post-training is famously different than other models. They target reliable operation of tiny models in ways other model families don't (they aren't just scaling a larger model to a smaller size). If you're looking for good performance out of tiny models, LFM has the most advanced design.

Note how they're much smaller than all other models in the comparison yet match or exceed them. This is for 2.6B params, but they have models as small as 230M. Nobody else designs models that small.

• woadwarrior01

  today at 6:40 AM

  > Note how they're much smaller than all other models in the comparison yet match or exceed them.

  There's a strong incentive to cherry pick in self-reported comparisons. If there is a model that's better, it gets left out. Have you seen Nanbeige4.2-3B or Ling-3.0-tiny?

  > Nobody else designs models that small.

  There are people building even smaller models.