Despite the success of therapies in lung cancer, more studies of new biomarkers for patient selection are urgently needed. The present study aims to analyze the role of galectin‐3 (GAL‐3) in the lung tumor microenvironment (TME) using tumorspheres as a model and explore its potential role as a predictive and prognostic biomarker in non‐small cell lung cancer (NSCLC) patients. For in vitro studies, lung adenocarcinoma (LUAD) and lung squamous carcinoma (LUSC) primary cultures from early‐stage patients and commercial cell lines were cultured, using tumorsphere‐forming assays and adherent conditions for the control counterparts. We analyzed the pattern of secretion and expression of GAL‐3 using reverse transcription–quantitative real‐time PCR (RTqPCR), immunoblot, immunofluorescence, flow cytometry and immunoassay analysis. Our results using three‐dimensional (3D) models of lung tumor cells revealed that soluble GAL‐3 (sGAL‐3) is highly expressed and secreted. To more accurately mimic the TME, a co‐culture of tumorspheres and fibroblasts was used, revealing that GAL‐3 could be important as an immunomodulatory molecule expressed and secreted in the TME, modulating immunosuppression through regulatory T cells (TREGS). In the translational phase, we confirmed that patients with high expression levels of GAL‐3 had more TREGS, which suggests that tumors may be recruiting this population through GAL‐3. Next, we evaluated levels of sGAL‐3 before surgery in LUAD and LUSC patients, hypothesizing that sGAL‐3 could be used as an independent prognostic biomarker for overall survival and relapse‐free survival in early‐stage LUAD patients. Additionally, levels of sGAL‐3 at pretreatment and first response assessment from plasma to predict clinical outcomes in advanced LUAD and LUSC patients treated with first‐line pembrolizumab were evaluated, further supporting that sGAL‐3 has a high efficiency in predicting durable clinical response to pembrolizumab with an area under curve (AUC) of 0.801 (p=0.011). Moreover, high levels might predict decreased progression‐free survival and overall survival to anti‐PD‐1 therapy, with sGAL‐3 being a prognosis‐independent biomarker for advanced LUAD.