FrontierThe story, in brief
Liquid AI Releases LFM2.5-VL-3B: A 3B Vision-Language Model That Reads Screens, Grounds Objects, and Calls Tools On-Device
3B vision-language model hits 87.9 on object grounding, runs natively on Apple Silicon at 228 tokens/s.
The KeyNews take
Why it matters
Liquid AI's LFM2.5-VL-3B demonstrates that capable multimodal reasoning (screen reading, object detection, tool calling) is now possible in 3B parameters on consumer devices. This shifts the frontier toward on-device deployment—practitioners can reason about visual context and call functions without cloud latency or API costs.
The key facts
6 to know3.1B parameters, ~3 GB footprint
ScreenSpot-v2: 80.7 average
RefCOCO grounding: 87.9 (up from 57.1)
ToolSandbox function calling: 59.5 (up from 26.4)
228 tokens/s on Apple M5 Max
Vision-language model with tool-calling capability
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Liquid AI released LFM2.5-VL-3B, a 3.1B-parameter vision-language model built for on-device deployment. It averages 80.7 on ScreenSpot-v2 and lifts RefCOCO grounding from 57.1 to 87.9. Function calling is new to the VL line, with ToolSandbox moving from 26.4 to 59.5. The model fits in roughly 3 GB…