Text-to-SQL generation enables non-experts to interact with databases vianatural language. Recent advances rely on large closed-source models like GPT-4that present challenges in accessibility, privacy, and latency. To addressthese issues, we focus on developing small, efficient, and open-sourcetext-to-SQL models. We demonstrate the benefits of sampling multiple candidateSQL generations and propose our method, MSc-SQL, to critique them usingassociated metadata. Our sample critiquing model evaluates multiple outputssimultaneously, achieving state-of-the-art performance compared to otheropen-source models while remaining competitive with larger models at a muchlower cost. Full code can be found at https://github.com/layer6ai-labs/msc-sql.