Imagine having an AI assistant that could quickly summarize key aspects of dense climate change reports. That's the promise of aspect-based summarization (ABS), a technique that allows users to extract information relevant to specific topics. Recent research has shown that not only large language models (LLMs) can perform this task, but smaller, more efficient models, known as small language models (SLMs), can achieve comparable results with a significantly reduced carbon footprint. This opens up exciting possibilities for making climate information more accessible, especially in resource-constrained settings. Researchers investigated the performance of various LLMs and SLMs on ABS using the newly released SumIPCC dataset, which contains summaries and corresponding paragraphs from IPCC climate change reports. Surprisingly, SLMs performed almost as well as their larger counterparts. While large LLMs like ChatGPT and GPT-4 achieved slightly higher scores on metrics like coherence and fluency, the difference wasn't statistically significant. Moreover, when factoring in energy consumption, the smaller models often outperformed larger models due to their efficiency. This suggests that SLMs are a viable and sustainable solution for summarizing climate information. The study also explored a more challenging scenario called Retrieval Augmented Generation (RAG), where the AI must first find the relevant paragraphs within the report before summarizing them. This mimics how someone might use the tool in the real world. While this task proved harder for all models, certain LLMs, particularly Llama 3, demonstrated a better ability to handle the added complexity. Interestingly, even in this scenario, smaller models like Qwen 1.8B held their own against the larger, more powerful models, showing potential for efficient RAG-based summarization. The SumIPCC dataset offers a benchmark for evaluating ABS systems in the climate domain, and opens avenues for future research. For example, further work could investigate how to incorporate more detailed information, such as section titles, during the retrieval process for RAG summarization. Additionally, fine-tuning SLMs on small, specific climate datasets could further enhance their performance. This research demonstrates the increasing capability of small AI models and their potential to provide crucial support in addressing complex issues like climate change. It’s a big step towards democratizing access to critical information and empowering individuals and organizations to make informed decisions.