Google’s Gemini 4 Argon has secured the top position in the first full round of the DualEntry Accounting AI Benchmark, achieving an industry-leading 91.1% accuracy score across 18 leading artificial intelligence models. Evaluated on its capability to handle complex financial and operational accounting tasks, the model also proved to be the most consistent platform tested, successfully passing 86.1% of tasks across three consecutive evaluation runs.
Google's Gemini 4 Argon achieved the highest accuracy (91.1%) and consistency (86.1%) among 18 models in the DualEntry Accounting AI Benchmark.
Zhipu AI's GLM-5.3 Flash and SpaceXAI's Grok 4.7 tied for second place with an 88.1% accuracy score.
Argon led or tied in five of eight accounting categories, posting perfect scores in transaction classification, financial reporting, and accounting knowledge.
Operational bottlenecks like accounts receivable and month-end close proved to be the toughest categories, averaging a 64.4% success rate across all models.
Cost variations were pronounced, with Claude Fable 5.1 costing $1.18 per task for an 85.1% score, while DeepSeek V4.1 Flash scored 84.2% at just $0.003 per task.
The evaluation used DualEntry's deterministic DualBench100 methodology, running 101 synthetic accounting cases three times without utilizing any customer data.
As finance teams increasingly look to autonomous AI agents for tasks ranging from routine data entry to period-end closes, consistency remains a critical hurdle. Benedict Dohmen, co-founder of DualEntry, noted that a model that gets a journal entry right once and wrong the next time is not ready for the books. While models like Grok 4.7 matched competitors on headline accuracy, consistency gaps caused their multi-run scores to fluctuate significantly more than Argon's.
Gemini 4 Argon’s success was driven by broad competency across multiple domains rather than isolated excellence. It achieved 100% scores in transaction classification, financial reporting, and accounting knowledge, and tied for the best result on hard tasks at 82.1%. However, the model still exhibited minor operational blind spots. Its weakest category was bank reconciliation at 75.0%—where Meta’s Muse Wonder led the field at 91.7%—while Moonshot AI’s Kimi K3 led the month-end close category at 91.7% compared to Argon’s 83.3%.
About DualEntry
DualEntry is the agentic, AI-native ERP for mid-market finance teams. It brings the general ledger, accounts payable and receivable, revenue recognition, multi-entity consolidation and reporting into one platform, with AI agents that take on accounting work from data entry through close. Founded in 2024 and based in New York, DualEntry raised a $90 million Series A in 2025 from Lightspeed Venture Partners, Khosla Ventures and GV.
