Open-Source AI Models Are Catching up to Closed Ones on Accounting Work, DualEntry Benchmark Finds

Open-Source AI Models Are Catching up to Closed Ones on Accounting Work, DualEntry Benchmark Finds

Moonshot AI's Kimi K3.1 Pro got 86.1% of tasks right on the DualEntry Accounting AI Benchmark, placing fourth and ahead of every model from Anthropic, OpenAI and Meta. With Zhipu AI's GLM-5.3 Flash tied for second at 88.1%, two of the top four models are now open.

The gap at the top is three points. Google's Gemini 4 Argon leads at 91.1%, and the best open model trails it by 3.0 points. Across the leaderboard, the seven open models average 83.9%, while the 13 closed models average 81.3%.

Key results:

• Kimi K3.1 Pro: 86.1%, fourth overall, up from 84.2% for Moonshot's previous model, Kimi K3
• Best open model: GLM-5.3 Flash at 88.1%, tied for second with Grok 4.7
• Open models hold four of the top seven places
• Bank reconciliation: Kimi K3.1 Pro scored 100%, against 75% for overall leader Gemini 4 Argon
• Cost: DeepSeek V4.1 Flash (open) scored 84.2% at $0.003 per task. Claude Fable 5.1 scored 85.1% at $1.18

Closed models still lead on reliability. The top score on hard tasks, 82.1%, is shared by three closed models, and the five models that most often passed a task in all three runs are closed. Gemini 4 Argon leads that measure at 86.1%.

"Finance teams have assumed that choosing an open model meant trading accuracy for cost and control," said Benedict Dohmen of DualEntry. "On accounting work that trade is mostly gone. What's left is consistency, and that is the gap to watch."

The DualEntry Accounting AI Benchmark (DualBench100) tests models on 101 synthetic accounting cases across eight areas: transaction classification, journal entries, financial reporting, bank reconciliation, accounts payable, accounts receivable, month-end close and accounting knowledge. The cases split into 37 easy, 36 medium and 28 hard. Each case has a fixed reference answer and is scored by deterministic rules with an 80% pass mark. Every model runs the full set three times. A task counts as passed when the model passes it in at least two runs. Full results are at dualentry.com/benchmark.

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. Learn more at dualentry.com.

Media contact
Ana Marturet anam@dualentry.com

Reporters can request task-level results, example questions and answers, or a call with the benchmark team.

View source version on businesswire.com: https://www.businesswire.com/news/home/20261026353736/en/