Abstract: How the recent progress of reasoning large language models (LLMs), especially the new open-source model DeepSeek-R1, can benefit financial services is an underexplored problem. While LLMs have ignited numerous applications within the financial sector, including financial news analysis and general customer interactions, DeepSeek-R1 further unlocks the advanced reasoning ability with multiple reinforcement learning-integrated training steps for more complex financial queries and provides distilled student models for resource-constrained scenarios. In this paper, we first introduce the technological preliminaries of DeepSeek-R1. Subsequently, we benchmark the performance of DeepSeek-R1 and its distilled students on two public financial question–answer (QA) datasets as a starting point for interdisciplinary research on financial artificial intelligence (AI). Then, we discuss the opportunities that DeepSeek-R1 offers to current financial services, its current limitations, and three future research directions. In conclusion, we argue for a proper approach to adopting reasoning LLMs for financial AI.

DeepSeek-R1遇上金融应用：基准、机遇与局限

[1]Achiam J, Adler S, Agarwal S, et al., 2023. GPT-4 technical report. https://arxiv.org/abs/2303.08774

[2]Bai JZ, Bai S, Chu YF, et al., 2023. Qwen technical report. https://arxiv.org/abs/2309.16609

[3]Bai YT, Jones A, Ndousse K, et al., 2022. Training a helpful and harmless assistant with reinforcement learning from human feedback. https://arxiv.org/abs/2204.05862

[4]Besta M, Blach N, Kubicek A, et al., 2024. Graph of thoughts: solving elaborate problems with large language models. Proc 38^th AAAI Conf on Artificial Intelligence, p.17682-17690.

[5]Brown TB, Mann B, Ryder N, et al., 2020. Language models are few-shot learners. Proc 34^th Int Conf on Neural Information Processing Systems, Article 159.

[6]Chang YP, Wang X, Wang JD, et al., 2024. A survey on evaluation of large language models. ACM Trans Intell Syst Technol, 15(3):39.

[7]Chervonyi Y, Trinh TH, Olšák M, et al., 2025. Gold-medalist performance in solving Olympiad geometry with AlphaGeometry2. https://arxiv.org/abs/2502.03544

[8]Daiya D, Lin C, 2021. Stock movement prediction and portfolio management via multimodal learning with Transformer. Proc IEEE Int Conf on Acoustics, Speech and Signal Processing, p.3305-3309.

[9]Fan T, Kang Y, Ma GQ, et al., 2023. FATE-LLM: an industrial grade federated learning framework for large language models. https://arxiv.org/abs/2310.10049

[10]Gemini Team of Google, 2023. Gemini: a family of highly capable multimodal models. https://arxiv.org/abs/2312.11805

[11]Guo DY, Yang DJ, Zhang HW, et al., 2025. DeepSeek-R1: incentivizing reasoning capability in LLMs via reinforcement learning. https://arxiv.org/abs/2501.12948

[12]Guo X, Xia HT, Liu ZW, et al., 2023. FinEval: a Chinese financial domain knowledge evaluation benchmark for large language models. https://arxiv.org/abs/2308.09975

[13]Hinton G, Vinyals O, Dean J, 2015. Distilling the knowledge in a neural network. https://arxiv.org/abs/1503.02531

[14]Hua WY, Zhang YF, 2022. System 1+system 2=better world: neural-symbolic chain of logic reasoning. Proc Findings of the Association for Computational Linguistics, p.601-612.

[15]Huang L, Yu WJ, Ma WT, et al., 2025. A survey on hallucination in large language models: principles, taxonomy, challenges, and open questions. ACM Trans Inform Syst, 43(2):42.

[16]Jaech A, Kalai A, Lerer A, et al., 2024. OpenAI o1 system card. https://arxiv.org/abs/2412.16720

[17]Kaplan J, McCandlish S, Henighan T, et al., 2020. Scaling laws for neural language models. https://arxiv.org/abs/2001.08361

[18]Kojima T, Gu SS, Reid M, et al., 2022. Large language models are zero-shot reasoners. Proc 36^th Int Conf on Neural Information Processing Systems, Article 1613.

[19]Kuang WR, Qian BC, Li ZT, et al., 2024. FederatedScope-LLM: a comprehensive package for fine-tuning large language models in federated learning. Proc 30^th ACM SIGKDD Conf on Knowledge Discovery and Data Mining, p.5260-5271.

[20]Lightman H, Kosaraju V, Burda Y, et al., 2024. Let’s verify step by step. Proc 12^th Int Conf on Learning Representations, p.1-24.

[21]Liu AX, Feng B, Xue B, et al., 2024. DeepSeek-V3 technical report. https://arxiv.org/abs/2412.19437

[22]Liu Y, Yao YS, Ton JF, et al., 2023. Trustworthy LLMs: a survey and guideline for evaluating large language models’ alignment. https://arxiv.org/abs/2308.05374

[23]Lusardi A, Mitchell OS, 2014. The economic importance of financial literacy: theory and evidence. J Econ Liter, 52(1):5-44.

[24]Miao XP, Oliaro G, Zhang ZH, et al., 2024. SpecInfer: accelerating large language model serving with tree-based speculative inference and verification. Proc 29^th ACM Int Conf on Architectural Support for Programming Languages and Operating Systems, p.932-949.

[25]Peng YF, Malin BA, Rousseau JF, et al., 2025. From GPT to DeepSeek: significant gaps remain in realizing AI in healthcare. J Biomed Inform, 163:104791.

[26]Qian LF, Zhou WP, Wang Y, et al., 2025. Fino1: on the transferability of reasoning enhanced LLMs to finance. https://arxiv.org/abs/2502.08127

[27]Shao ZH, Wang PY, Zhu QH, et al., 2024. DeepSeekMath: pushing the limits of mathematical reasoning in open language models. https://arxiv.org/abs/2402.03300

[28]Sun GY, Jin MY, Wang ZT, et al., 2024. Visual agents as fast and slow thinkers. https://arxiv.org/abs/2408.08862

[29]Touvron H, Martin L, Stone K, et al., 2023. Llama 2: open foundation and fine-tuned chat models. https://arxiv.org/abs/2307.09288

[30]Vaswani A, Shazeer N, Parmar N, et al., 2017. Attention is all you need. Proc 31^st Int Conf on Neural Information Processing Systems, p.6000-6010.

[31]Webb T, Holyoak KJ, Lu HJ, 2023. Emergent analogical reasoning in large language models. Nat Human Behav, 7(9):1526-1541.

[32]Wei J, Wang XZ, Schuurmans D, et al., 2022. Chain-of-thought prompting elicits reasoning in large language models. Proc 36^th Int Conf on Neural Information Processing Systems, Article 1800.

[33]Wu SJ, Irsoy O, Lu S, et al., 2023. BloombergGPT: a large language model for finance. https://arxiv.org/abs/2303.17564

[34]Yao SY, Yu D, Zhao J, et al., 2023. Tree of thoughts: deliberate problem solving with large language models. Proc 37^th Int Conf on Neural Information Processing Systems, Article 517.

[35]Zelikman E, Harik G, Shao Y, et al., 2024. Quiet-STaR: language models can teach themselves to think before speaking. https://arxiv.org/abs/2403.09629

[36]Zhang WT, Zhao LX, Xia HC, et al., 2024. A multimodal foundation agent for financial trading: tool-augmented, diversified, and generalist. Proc 30^th ACM SIGKDD Conf on Knowledge Discovery and Data Mining, p.4314-4325.

[37]Zhang XY, Yang Q, 2023. XuanYuan 2.0: a large Chinese financial chat model with hundreds of billions parameters. Proc 32^nd ACM Int Conf on Information and Knowledge Management, p.4435-4439.

[38]Zhou J, Ke P, Qiu XP, et al., 2024. ChatGPT: potential, prospects, and limitations. Front Inform Technol Electron Eng, 25(1):6-11.