LFM
No written verdict yet. Here is what commenters said.
8recommend
0warn against
Jan 10, 2026first mention
Oct 27, 2026latest mention
Why people recommend it
“Excited to roll out LFM 3 to replace it, looks like a big step up.”
switched from LFM 2.6on “Introducing LFM3: Every Modality for Specialized Int…”
“I think the d1-2B is amazing”
Show 3 more
“i chose lfm over bert as the base model because it uses grouped-query attention”
“Tell me your username and i'll give you (six) URLs”
“these LFM 2.5 models are crazy fast.”
Caveats and dissent
No one in these comments warned against it.
Threads where it came up
- Introducing LFM3: Every Modality for Specialized Intelligence 3 mentions
- Show HN: Offlineisbetter: Efficient Models for Text on CPU 1 mention
- RSS Feeds for Last.fm 1 mention
- Liquid AI releases a 230M model optimized for phones, Raspberry Pi, and robots 1 mention
- LFM2-24B-A2B: Scaling Up the LFM2 Architecture 1 mention
- Local LLM by Ente 1 mention
Also written as LFM2, LFM 2.5, LFM2.5, LFM3. Data: lfm.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 10
“fine-tuned an LFM encoder for promptable classification”
“Excited to roll out LFM 3 to replace it, looks like a big step up.”
“I think the d1-2B is amazing”
“i chose lfm over bert as the base model because it uses grouped-query attention”
“Tell me your username and i'll give you (six) URLs”
“these LFM 2.5 models are crazy fast.”
“Liquid AI have made some awesome models (especially the smaller ones, they are lightning fast).”
“the LFM model license has a commercial use threshold of $10m”
“the LFM2 models run pretty fast on a phone and are surprisingly capable”
“for edge running look into the LFM 2.5 class of models”