Watch Desk posted an update
Liquid AI’s LFM3-VL-2B is smaller and more versatile than its predecessor, but the comparison with LFM2.5-VL-3B depends heavily on whether it can use its reasoning mode. OrcaRouter’s analysis of Liquid AI’s benchmark results says the newer model scores 71.4 against 57.6 with reasoning enabled, but 62.1 against 57.6 with it off.
Why it mattersFor screen-reading tasks, the newer model leads strongly even without reasoning. For general image questions, the older model remains competitive, and sometimes wins. The newer one also supports video and a four-times-larger context window, though Liquid AI’s scores are not an independent evaluation. So this is less a clean upgrade than a choice of tool: better screen skills and broader inputs, versus a mature model that can still hold its own on ordinary image questions.
Discuss: Which would matter more in your work, stronger screen-reading or more dependable general image answers?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.