Databricks Launches AI Index Tracking Enterprise Model Adoption, Spending and Coding Agent Cost-Quality Trade-offs
New public index from 10,000+ Unity Gateway customers shows open-weight model token share quadrupled to 31%, single-provider reliance fell from 64% to 28%, and token prices don't predict coding agent costs.
By TechWhoKnows
6 min read
Databricks has launched the Databricks AI Index, a public resource tracking real-world enterprise AI model adoption, spending patterns and coding agent performance across more than 10,000 companies using Unity Gateway.
The index combines two data sources: aggregated, anonymised usage and spending data from Unity Gateway customers (spanning digital natives, Fortune 500 and Global 2000 enterprises), and continuously updated coding-agent evaluations drawn from real workflows by Databricks engineers.
Enterprise model preferences shifting rapidly
At mid-2026, most organisations relied on a handful of Anthropic models. By November, adoption fragmented significantly across model versions and providers. OpenAI's share of enterprise token traffic grew meaningfully, driven by its 6 series Sol and Astra models. However, model quality still strongly influences provider share: when Claude Opus 5.5 launched, it surged from 7.2% to 37.9% of proprietary-model tokens in six weeks, lifting Anthropic's overall share from 57.6% to 68.8%.
Databricks uses an internal Day 1 model rollout process to evaluate new releases within days. This helped the company find that Opus 5.5 delivered improved quality at 29% lower cost per session than its predecessor, informing the decision to make it the default model for Claude Code.
Multi-provider and open-weight adoption accelerating
Enterprises are increasingly moving beyond a single AI provider. In July, approximately 64% of Unity Gateway enterprises relied on just one model provider. By 2 November, that had fallen to 28% as more organisations adopted models from two or more providers.
Open-weight models are gaining ground alongside this shift. Their share of enterprise token traffic grew from 7.3% in July to 30.9% in November, more than quadrupling in less than four months. In Databricks' coding quality benchmark, three open-weight models rank among the top 10, including DeepSeek V4.1 and GLM 5.3, both at 93.1% pass rate.
No single model or harness dominates the cost-quality frontier
The latest coding-agent benchmark evaluates models across nine AI labs using Claude Code and Codex. Five models from four labs (OpenAI, DeepSeek, Meta and Anthropic) sit on the cost-quality frontier, delivering the best trade-offs between task success rates and cost.
Databricks uses this approach internally. When evaluating whether to upgrade from Opus 4.8 to Opus 5, the team analysed the cost-quality Pareto frontier and replayed traces from their own workloads. The upgrade was not worth the additional cost for their use cases, so they stayed on 4.8.
Token prices don't tell the full story of coding costs
Two models can deliver nearly the same benchmark performance at dramatically different costs. In the coding benchmark, GPT-6 Luna and Claude Opus 5.5 both achieved an 89.1% task success rate. However, Luna averaged just $0.05 per task compared with $1.62 for Opus 5.5, a roughly 32x difference.
Cost per task depends on both token prices and the amount of work a model must do to reach an answer. Extra turns, long outputs and failed attempts all increase cost. Gemini 3.8 Flash has a low token price yet averaged 140 turns per task and finished as one of the most expensive models at $11.97 per task.
Highest-performing model is rarely the best value
Claude Fable 5.1 tops the benchmark with a 96.0% task success rate at $7.47 per task, while Muse Wonder scores 94.9% at just $1.53. That additional 1.1 percentage points of performance comes at nearly five times the cost. Enterprises appear to agree: Fable 5.1's share of proprietary tokens fell from 5.5% to 0.6% over six weeks since its launch.
AI spending remains highly concentrated
The top 10% of users account for 78% of AI spend. Median weekly spend per active user is $0.87, while the 90th percentile spends $12.40. This concentration suggests most enterprises are still in early adoption phases with a small number of power users driving the majority of value.
Unity Gateway integration
Databricks is making these insights actionable through Unity Gateway features like Smart Routing, which can match tasks to the most cost-effective model-harness combination based on real-time index data.
Sources
- 1Databricks Blog - Introducing the Databricks AI Indexdatabricks.com