DeepSeek V4.1 Flash 1017 Declines in Accounting Benchmark as Closed Models Maintain Reliability Lead

Yahoo Finance · 2026-10-21

According to the DualEntry Accounting AI Benchmark released on October 21, 2026, DeepSeek's V4.1 Flash 1017 achieved an accuracy rate of 82.2%, dropping from 84.2% for its predecessor. The updated version is tied for tenth among 19 models alongside Zhipu AI's GLM-5.3 and Anthropic's Claude Opus 5.5. Google's Gemini 4 Argon led the assessment with 91.1%, while Zhipu AI's GLM-5.3 Flash emerged as the top open model at 88.1%. DualEntry founder Benedict Dohmen noted that while open-source models have nearly closed the accuracy gap, proprietary closed models continue to dominate in reliability and execution consistency across multiple runs, which is critical for automated financial reporting and ledger management.

AI 시장 분석

Dual-entry accounting AI benchmark results showed DeepSeek's latest model, V4.1 Flash 1017, recording an accuracy of 82.2%, a decline from its previous version. Closed models such as Google's Gemini 4 Argon dominated the top ranks with 91.1% accuracy and high consistency. Open-source models have narrowed the gap in accuracy, but still lag in reliability and consistency, showing limitations in entering the enterprise accounting market.

상승 영향

AI — Providers of closed models, such as Google, proved exceptional consistency and accuracy in accounting and complex tasks, solidifying their technological dominance in the enterprise AI market.

하락 영향

Software — Open-source AI models exposed a lack of consistency and reliability in professional tasks like accounting, putting the brakes on their penetration into the enterprise software market.

DYAX 전담 분석

In accounting benchmark tests, open-source models exposed a lack of consistency and reliability, raising concerns about weakened competitiveness in the enterprise automation market. This suggests that the moat of closed models remains robust in terms of the stability required for practical deployment, going beyond simple accuracy-centered evaluations.

In the future enterprise AI market, consistency metrics during repeated execution are expected to emerge as a core competitive edge over one-off performance indicators. Investors should pay attention to the pace of reliability improvement in the open-source camp and the expansion of market share by companies holding closed models.

AI가 생성한 분석으로 투자 자문이 아닙니다.

DYAX Investor Sentiment

Bullish (Long) 50% · Bearish (Short) 50%

549 participants

Related News

원문 보기 — Yahoo Finance