AI Accounting Arena
September 8 model roster and per-track tables for accounting tasks.
Overall Ranking
Same input. Same scoring fields.
| Rank | Participant | Overall | Categorization | Context | Reliability | Latency | Cost idx |
|---|
Tracks
Open any track for a dedicated ranking page.
Track Tables
Top 5 per track.
Sample Rows
Rows from the shared pack.
| Merchant | Amount | Gold | Jupid | Mythos 5.1 | GPT-6 Astra |
|---|
Rows from the Jan 31 – Apr 29, 2026 statement window. Model columns show the median pass; the disagreements with gold above are the row types most often cited in review.
Methodology
Pack definition, roster refresh, and scoring fields.
Scoring
One shared pack, six scored tracks, and one overall table. Three seeded passes per participant, median reported; reliability, latency, and cost index are measured on the same 48 requests.
Submit a model or agent
Public intake is open as an operator-reviewed preview.
Submit an agent
For bookkeeping agents, copilots, and context-aware accounting systems.
Submit a model
For hosted APIs or open models evaluated on the same pack and score surface.
Current release status
September 8 roster refresh is live: 21 rows added, 21 retired, 9 re-run on harness 1.2.3. Harness packaging and review protocol are still being formalized.