---
title: "How Databricks rolls out frontier models to 14,000 employees on Day 1"
description: "Databricks provides employees Day 1 access to new AI models using Unity Gateway for governance and cost management. The company uses a three-step pipeline: immediate experimental availability, per-user budget constraints, and data-driven promotion decisions. Unity Gateway CLI distributes model configurations to employee laptops, tagging new models as experimental. Per-user budgets include a monthly maximum, daily runaway limit, quality frontier budget, and experimental budget to control costs. Within three days of the Opus 5.5 and GPT-6 Sol launches, Databricks gathered benchmark, user feedback, and cost data to confirm they were on the efficiency frontier and promoted them, with Opus 5.5 showing a 29% cost reduction per session over Opus 4.8 and GPT-6 Sol showing a 48% reduction over GPT-5.6 Sol."
---

# How Databricks rolls out frontier models to 14,000 employees on Day 1

[Databricks](https://yomu.fyi/company/databricks) · The Databricks AI Product and Engineering Team · Sep 28, 2026

**Type:** Problem & solution

## Summary

Databricks provides employees Day 1 access to new AI models using Unity Gateway for governance and cost management. The company uses a three-step pipeline: immediate experimental availability, per-user budget constraints, and data-driven promotion decisions. Unity Gateway CLI distributes model configurations to employee laptops, tagging new models as experimental. Per-user budgets include a monthly maximum, daily runaway limit, quality frontier budget, and experimental budget to control costs. Within three days of the Opus 5.5 and GPT-6 Sol launches, Databricks gathered benchmark, user feedback, and cost data to confirm they were on the efficiency frontier and promoted them, with Opus 5.5 showing a 29% cost reduction per session over Opus 4.8 and GPT-6 Sol showing a 48% reduction over GPT-5.6 Sol.

## Context

Databricks needed to give over 10,000 employees rapid access to new frontier AI models while avoiding quality regressions and cost explosions.

## Approach / What changed

Databricks uses a three-step model release lifecycle: immediately make new models available as experimental via Unity Gateway, constrain usage with per-user budgets (monthly max, daily runaway, quality frontier, experimental), and after gathering benchmark, user feedback, and cost data within days, promote or drop the model.

## Takeaways

- Naive model rollout can cause a 60% cost increase, as seen when GPT Astra was released without cost mitigations.
- Databricks uses stratified cost analysis (reweighting sessions by single/multi-turn and file edits) to fairly compare model costs per session.
- Within three days of launch, Databricks confirmed Opus 5.5 and GPT-6 Sol were on the efficiency frontier and promoted them, with Opus 5.5 becoming the default for Claude Code.

## FAQ

### How does Databricks prevent cost explosions from new model access?

Databricks uses per-user budgets including a monthly maximum, daily runaway limit, quality frontier budget, and experimental budget. The experimental budget allocates a fraction of monthly spend for untested models, balancing adoption speed with downside risk.

### What signals does Databricks use to decide whether to promote a model?

Databricks relies on three signals: private benchmark data (offline and online), user-reported quality from power users, and cost tracking via OpenTelemetry traces. These are gathered within days of experimental release.

### How does Databricks ensure fair cost comparison between old and new models?

Databricks stratifies sessions by single/multi-turn and file edits, then reweights the distribution to account for early adopters trying harder problems. This normalized $/session comparison revealed Opus 5.5 was 29% cheaper than Opus 4.8.

**Tags:** [AI](https://yomu.fyi/topic/ai), [Cost Optimization](https://yomu.fyi/topic/cost-optimization), [LLMs](https://yomu.fyi/topic/llm), [Model Evaluation](https://yomu.fyi/topic/model-evaluation), [Unity Gateway](https://yomu.fyi/topic/unity-gateway)

- Source: [Databricks](https://www.databricks.com/blog/how-databricks-rolls-out-frontier-models-14000-employees-day-1)
- Source URL: https://www.databricks.com/blog/how-databricks-rolls-out-frontier-models-14000-employees-day-1
- Ingested by Yomu: 2026-09-28T21:00:07.317Z

[Read original post](https://www.databricks.com/blog/how-databricks-rolls-out-frontier-models-14000-employees-day-1)