Back to News
ProductAI Understanding briefing

Databricks launches AI Index to track enterprise model adoption and cost

Databricks unveiled the AI Index, a public, weekly‑updated resource that aggregates anonymized usage data from thousands of enterprises and benchmarks coding‑agent performance across proprietary and open‑weight models.

4 min readRead the primary source
Source-provided image accompanying Databricks launches AI Index to track enterprise model adoption and cost
Primary-source documentSource recorded
Publisher
databricks.com
Source type
Primary document — an official announcement, paper, filing, or first-party page we read directly.
ContextUnderstand this in 60 seconds

Key terms

Benchmark
A standardized test or dataset used to measure and compare model performance.
Weight
A learned numeric value that scales signals passing through a neural network.
Token
A chunk of text processed by language models, such as a word piece or symbol.

What happened

Databricks announced the public launch of the Databricks AI Index, a resource that combines aggregated, anonymized usage data from over 10,000 enterprises using Unity Gateway with real‑world coding‑agent results. The Index tracks model adoption, and spend shares, and weekly spending patterns, and it will be refreshed weekly as new models are evaluated. The company also released a new internal “Day 1” model rollout process that lets Databricks evaluate new model releases within days, informing internal model selection and cost‑quality decisions.

Databricks introduced the AI Index as a publicly accessible website that aggregates anonymized data from more than 10,000 companies using its Unity Gateway platform. The data includes traffic, spend shares, and weekly spending patterns for both proprietary and open‑ models, covering the period from September 21 to November 7, 2026.

The Index also incorporates a coding‑agent that evaluates model‑harness combinations (e.g., Claude Code, Codex) on real pull‑request workflows drawn from Databricks engineers’ private codebase. Benchmarks measure task success rates and cost per completed task using hidden test suites to avoid contamination.

Key findings highlighted in the launch blog include a sharp fragmentation of model usage (OpenAI’s share grew, Anthropic’s Opus 5.5 surged to 37.9% of proprietary tokens within six weeks), a four‑fold rise in open‑ model token traffic (from 7.3% to 30.9% between July and November), and large cost disparities between models with similar performance (e.g., GPT‑6 Luna at $0.05 per task versus Claude Opus 5.5 at $1.62).

Databricks also described its internal “Day 1” model rollout process, which evaluates new releases within days, allowing the company to adopt higher‑quality, lower‑cost models quickly. This process informed the decision to make Opus 5.5 the default for Claude Code after finding it delivered 29% lower cost per session with improved quality.

The blog concludes with a call to action for enterprises to use Unity Gateway’s dashboards, unified traces, and Smart Routing to replicate the Index’s analysis on their own workloads, aiming to improve budgeting, cost control, and model selection.

Source details: databricks.com

Why it matters

The AI Index gives enterprises a rare, data‑driven view into how leading AI models are being used, how much they cost, and how performance varies across model‑harness combinations. By publishing anonymized usage from thousands of customers, Databricks provides a that can help organizations avoid costly trial‑and‑error when selecting models for coding agents, budgeting AI spend, and designing routing policies. The Index also highlights the rapid rise of open‑ models, showing they can compete with proprietary offerings on both quality and price—a shift that could reduce vendor lock‑in and reshape procurement strategies. For the broader AI ecosystem, the Index creates a transparent, continuously updated reference point for model tokenomics, encouraging more rigorous cost‑performance analysis and potentially prompting providers to adjust pricing or improve efficiency.

Transparency: By publishing aggregated usage and data, Databricks fills a gap in the market where enterprises typically rely on opaque vendor claims or limited internal testing.

Cost optimization: The Index quantifies the trade‑off between model performance and expense, showing that cheaper prices do not always equal lower total cost—a critical insight for budgeting AI‑driven development.

Vendor diversification: The rapid rise of open‑ models in the Index suggests enterprises are moving away from single‑provider lock‑in, which could pressure proprietary vendors to improve pricing or capabilities.

Strategic decision‑making: Enterprises can compare their own trace data against the Index’s cost‑quality frontier, enabling data‑driven choices about model upgrades, routing policies, and spend caps.

Industry : The Index may become a de‑facto reference for AI tokenomics, influencing how other cloud and AI platform providers report usage and cost metrics.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
Interactive Concept Check+10 Points
AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

What to watch next

Watch for how enterprises adopt the Index in their own governance workflows, whether model providers respond with pricing changes, and how the weekly updates influence the market share of emerging open‑ models. Also monitor Databricks’ Smart Routing adoption rates and any extensions of the Index to non‑coding workloads or other AI use cases.

Adoption rates of the AI Index within Databricks customers and whether it drives measurable spend reductions.

Responses from major model providers (OpenAI, Anthropic, Meta, etc.) to the public cost‑performance data, potentially leading to price adjustments or new model releases.

Expansion of the Index beyond coding agents to other enterprise AI workloads such as text generation, summarization, or analytics.

Evolution of Databricks’ Smart Routing and budget‑cap features as enterprises seek to operationalize the Index insights.

Emergence of competing industry‑wide tokenomics dashboards that could validate or challenge Databricks’ findings.

Related guides & quizzes

Found this useful?