Liquid AI Launches LFM2.5-VL-3B: A Lightweight Vision-Language Model Outperforming 2.6x Larger Rivals
JosephJacks_ · x · 2026-08-13
Liquid AI has released LFM2.5-VL-3B, a lightweight vision-language model. Built on the LFM2.5-2.6B base and a SigLIP2 vision encoder, it was pre-trained on 34T tokens.
The model specializes in reading screens, documents, and the physical world across mobile, web, and desktop platforms. It supports tool calling from both text and image inputs. In benchmarks, it delivers impressive results:
- ScreenSpot-v2: 80.7 (beating Gemma-4-E4B at 51.2)
- RealWorldQA: 73.1 (beating InternVL-3.5-4B at 67.7)
- TextVQA: 84.3 (beating Qwen3.5-4B at 81.2)
Achieving performance comparable to or better than models up to 2.6x its size, it serves as a highly efficient base for custom applications.
Related event: LiquidAI Releases LFM2.5-VL-3B Edge Multimodal Model(7 posts)→
More from Models
- Grok 4.6 Launch Draws Criticism Over Missing Model Card and Safety Tests — Miles_Brundage · 2026-08-13
- Upstage Solar Pro 4 Review: High Intelligence but Notably Slow — ArtificialAnlys · 2026-08-13
- Solar Pro 4's Lower Hallucination Rate Comes from Abstention, Not Knowledge — ArtificialAnlys · 2026-08-13
- AI Coding Benchmarks Under Fire: Secret Tests and Suspected Bias — astralmatrix · 2026-08-13
- Grok-4.6 Takes the Lead on CursorBench and FrontierCode — scaling01 · 2026-08-13
- Claude Opus 5 Tops InferenceBench with 8.9x Speedup Over PyTorch — maksym_andr · 2026-08-13