Databricks says Genie Code beat three coding agents on 401 tasks at $0.55 each
matei_zaharia · x · 2026-07-26
Databricks says better data agents can improve accuracy and efficiency at the same time.
In a head-to-head test on 401 real internal data tasks, Genie Code was compared with three leading general-purpose coding agents. Each agent used its own harness, frontier models, Databricks MCP, and the same 20-minute task budget.
Reported results:
- Accuracy: 76.6%, the highest of the group
- Mean cost: $0.55 per task, the lowest of the group
- Relative cost: less than half the cost of the other agents
The company says the edge came from context handling: semantic search, persistent memory, and workspace understanding reduced random exploratory wandering and helped agents arrive at the right answer faster.
Related event: Databricks Claims Genie Code Beats Coding Agents at $0.55/Task(2 posts)→
More from coding & agent
- NVIDIA says Nemotron 3 Ultra hit 97.1% on agentic RTL chip-design tasks — NVIDIAAI · 2026-07-27
- Tokyo Agent Forge hackathon shipped production-ready AI agents in one day — DavidBennett__ · 2026-07-27
- Long-running agents will need immutable event logs, this thread argues — sebpaquet · 2026-07-27
- Agentic Data Science in Practice: Agents Write Code but Answer Wrong Questions — hugobowne · 2026-07-27
- A VS Code extension adds Markdown-style highlighting to Alchemy string templates — samgoodwin89 · 2026-07-27
- Claude’s Stripe MCP connector is being called unusable after repeated disconnects — evielync · 2026-07-27