---
title: "How Databricks rolls out frontier models to 12,000 employees on Day 1"
description: "Databricks provides 12,000 employees Day 1 access to new frontier models through a pipeline using Unity Gateway for governance, cost management, and observability. New models are immediately made available on an experimental basis with per-user budgets to control costs and mitigate risk from models that may not be on the efficiency frontier. Budgets include monthly maximum, daily runaway limit, quality frontier budget, and experimental budget, with configurations pushed via Unity Gateway CLI running on employee laptops. After three days, Databricks evaluates models using benchmark data, user-reported quality, and cost tracking via OpenTelemetry traces to decide whether to promote the model. Opus 5.5 was promoted as higher quality and lower cost than predecessors, while GPT-6 Sol was included in the smart router due to its cost advantage over GPT-5.6 Sol."
---

# How Databricks rolls out frontier models to 12,000 employees on Day 1

[Databricks](https://yomu.fyi/company/databricks) · The Databricks AI Product and Engineering Team · Sep 29, 2026

**Type:** Explainer

## Summary

Databricks provides 12,000 employees Day 1 access to new frontier models through a pipeline using Unity Gateway for governance, cost management, and observability. New models are immediately made available on an experimental basis with per-user budgets to control costs and mitigate risk from models that may not be on the efficiency frontier. Budgets include monthly maximum, daily runaway limit, quality frontier budget, and experimental budget, with configurations pushed via Unity Gateway CLI running on employee laptops. After three days, Databricks evaluates models using benchmark data, user-reported quality, and cost tracking via OpenTelemetry traces to decide whether to promote the model. Opus 5.5 was promoted as higher quality and lower cost than predecessors, while GPT-6 Sol was included in the smart router due to its cost advantage over GPT-5.6 Sol.

## Context

Databricks needed to give 12,000 employees rapid access to new frontier models while avoiding quality regression (e.g., Opus 5.0 was worse than Opus 4.8) and cost explosions (e.g., GPT Astra caused a 60% cost increase). They developed a lifecycle using Unity Gateway for governance and evaluation.

## Approach / What changed

Databricks uses Unity Gateway with a three-step pipeline: immediately release new models on an experimental basis with per-user budgets, collect data over three days using benchmarks, user reports, and stratified cost tracking, then promote or drop the model based on efficiency frontier.

## Takeaways

- New models are released experimentally with per-user budgets to control costs and limit risk from models that may not be on the efficiency frontier.
- Databricks stratifies sessions by complexity (single/multi-turn, file edits) and reweights distributions to fairly compare model costs per session.
- Opus 5.5 was 29% cheaper per session than Opus 4.8 with higher quality, while GPT-6 Sol had a 48% cost reduction but did not replace the default model.

## FAQ

### How does Databricks control costs when rolling out new models?

Databricks uses per-user budgets including a monthly maximum, daily runaway limit, quality frontier budget, and experimental budget. The experimental budget limits exposure to potentially inferior models while allowing data collection.

### What signals decide whether to promote a new model?

They use three signals: private benchmark data (e.g., OfficeQA Pro V2), user-reported quality via surveys and Slack, and cost tracking from OpenTelemetry traces, stratifying sessions by complexity to compare costs fairly.

### How much cheaper was Opus 5.5 compared to Opus 4.8?

After stratified cost analysis, Opus 5.5 was 29% cheaper per session than Opus 4.8 while providing higher quality, leading to its promotion as a default model.

**Tags:** [AI Governance](https://yomu.fyi/topic/ai-governance), [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-12000-employees-day-1)
- Source URL: https://www.databricks.com/blog/how-databricks-rolls-out-frontier-models-12000-employees-day-1
- Ingested by Yomu: 2026-09-29T21:00:08.386Z

[Read original post](https://www.databricks.com/blog/how-databricks-rolls-out-frontier-models-12000-employees-day-1)