AIToday
AI Stocks & MarketsAI Business & IndustryDatabricks AI BlogPublished: Nov 17, 2026, 04:00 JST

Databricks AI Index: single-provider use falls to 28%

Send on LINE
Databricks AI Index: single-provider use falls to 28%

3 Key Points

  1. What happened

    Databricks launched the Databricks AI Index, a public resource drawing on anonymized usage data from 10,000+ companies on Unity Gateway plus weekly coding-agent benchmarks; in the first dataset, enterprises relying on one model provider fell from about 64% in July to 28% by November 2.

  2. Why it matters

    That sharp drop suggests enterprises are spreading AI workloads across multiple model providers rather than standardizing on one, making centralized access and spend visibility a more pressing need for the teams buying and governing these tools.

  3. What to watch

    The Index is updated weekly as new proprietary and open-weight models are evaluated, with the current dataset covering September 21 through November 7, 2026.

WHO IT HITSEnterprise IT and platform teams that procure and govern AI models now have a public, weekly benchmark for comparing adoption, quality and cost per task across providers, which may reshape how they justify model choices and budgets.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

The Databricks AI Index arrives as enterprises face an unusually crowded model market: since May 2026, a new flagship proprietary or open-weight model has launched roughly every week, or every 3.7 days when regional labs are included. Databricks says customers have repeatedly asked how new models compare and what adoption looks like across enterprises, and the Index makes its internal evaluation suite permanently public with weekly updates.

The first dataset points to a rapid fragmentation of enterprise AI usage. At mid-year, most organizations relied on a handful of Anthropic models for the majority of their work, but adoption has since spread across model versions and providers. Databricks also describes its own Day 1 model rollout process, which lets it evaluate new releases within days, and says this helped it manage cost while improving internal AI quality.

The Index also highlights a gap between benchmark performance and cost. In Databricks' coding benchmark, GPT-6 Luna and Claude Opus 5.5 both achieved an 89.1% task success rate, but Luna averaged $0.05 per task versus $1.62 for Opus 5.5, a roughly 32x difference. Databricks says enterprises can use Unity Gateway dashboards and traces to analyze model consumption and costs on their own workloads, and Smart Routing to match tasks to cost-effective models.

FAQ
What does the Databricks AI Index track?
It tracks enterprise AI model adoption through token and spend shares across proprietary and open-weight models, weekly AI spending patterns for individual users, and coding-agent evaluations based on real workflows from Databricks engineers.
How often is the Databricks AI Index updated?
The Index is updated weekly as new proprietary and open-weight models are evaluated, with the current dataset covering September 21 through November 7, 2026.
Do lower token prices always mean lower AI coding costs?
No. Cost per task depends on both token prices and how much work a model does to reach an answer; Gemini 3.8 Flash has a low token price but averaged 140 turns per task and cost $11.97 per task.
Databricks AI BlogRead Original Article
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleLean overtakes Isabelle as theorem-proving research shifts, survey of 208 papers shows