LiquidAI LFM2.5-1.2B-Thinking is a 1.2B parameter model from LiquidAI with a dedicated thinking mode, available at no cost. Despite the reasoning orientation, benchmark performance across intelligence, coding, and instruction following is very limited at this scale. The 32K context window and text-only modality further constrain its utility.

For businesses, the thinking mode adds marginal value over the standard instruct variant for simple structured problems, but the capability ceiling remains very low. It is not suited for any task requiring reliable multi-step reasoning, coding, or nuanced output at a professional standard.

Being free makes it worth testing for bulk preprocessing or simple classification pipelines, but most business workloads will require a substantially more capable model. It functions best as a low-cost filter in a tiered routing architecture rather than a standalone tool.

Assessed June 17, 2026

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?