Databricks launched the AI Index tracking tokens and spend across 10,000+ enterprises using Unity Gateway.
Since May 2026 a new flagship model has shipped roughly every week, per Databricks. Regional labs push the pace to every 3.7 days, per Databricks. Databricks said customers keep asking which model wins on quality and cost.
Loyalty is collapsing.
In six weeks Claude Opus 5.5 jumped from 7.2% to 37.9% of proprietary tokens, per Databricks. Anthropic’s share rose from 57.6% to 68.8% in that span, per Databricks. OpenAI had gained earlier on its 6 series Sol and Astra models, per Databricks.
Opus 5.5 is available natively on the Claude Platform and through AWS, Google Cloud and Microsoft Foundry, per Anthropic.
Single-provider enterprises fell from 64% in July to 28% by November 2, per Databricks. Open-weight share more than quadrupled from 7.3% to 30.9%, per Databricks. DeepSeek V4.1 and GLM 5.3 now hit 93.1% pass rates in coding tests, per Databricks.
GPT-6 Luna and Opus 5.5 both hit 89.1% task success, per Databricks. Luna averaged $0.05 per task against $1.62 for Opus 5.5, per Databricks. Gemini 3.8 Flash needed 140 turns and cost $11.97 per task, per Databricks. Token price misleads when harnesses run hot.
Claude Fable 5.1 leads at 96.0% for $7.47, while Muse Wonder hits 94.9% for $1.53, per Databricks. Fable’s token share fell from 5.5% to 0.6% in six weeks, per Databricks. Median weekly spend was $2.49 against $699.74 at the 99th percentile, per Databricks.
Databricks will update the Index weekly with new models and usage data. The current set covers September 21 through November 7.





