LiquidAI: LFM2.5-1.2B-Thinking (free)

LiquidAI LFM2.5-1.2B-Thinking is a text-only model from Liquid AI with a 32,768-token context window. It supports reasoning, meaning it can work through multi-step problems before producing a response, but it does not support tool use or structured output. The absence of tool calling and structured output limits its fit for agentic pipelines or applications that require reliably formatted responses. The model is free to use, which makes it worth a look for developers prototyping on a budget or researchers who want to experiment with reasoning-capable models at no cost. The significant caveat is that there is no independent benchmark coverage yet, so its actual performance relative to other reasoning models is unproven. Shortlisting it makes most sense when cost is the primary constraint and you can evaluate quality directly on your own tasks, rather than relying on published scores.