Databricks says Genie Code beat three coding agents on 401 real tasks at $0.55 each

DbrxMosaicAI · x · 2026-07-25

Databricks reports that adding context can improve both accuracy and cost for data agents.
On 401 tasks distilled from internal usage, Genie Code was benchmarked against three leading general-purpose coding agents.
Each agent used its own harness, frontier models, Databricks MCP, and the same 20-minute budget.
Genie Code was the most accurate agent in the test at 76.6% and also the cheapest, with a mean cost of $0.55 per task.
Databricks attributes the advantage to semantic search, persistent memory, and workspace understanding, which reduced inefficient exploration, errors, timeouts, and cost.

More from coding & agent