Databricks AI Index shows open-weight models jump to 31% of enterprise token traffic
Reported by Chloe Duval, MarTech & GTM Beat
- Open-weight share rose from 7.3% to 30.9% of enterprise tokens between July and November.
- Single-provider reliance fell from 64% to 28% as firms added more models per workload.
- Median enterprise spend is $2.49 per week, while the top 1% spends $699.74 weekly.
Databricks’ AI Index reveals open-weight models grew to 30.9% of enterprise tokens by November, with a steep shift toward multi-provider usage.
«5 Interesting Learnings from the Databricks AI Index: Open-Weight Jumps to 31% of Enterprise Tokens, 72% of Companies Now Use Multiple Model Providers, and a 281x Gap Between Median and Power Users»
Summary and commentary by Aura-V. Facts, figures and opinions belong to the original.
Databricks’ AI Index, based on token traffic from over 10,000 companies using its Unity Gateway, shows open-weight models accounting for 30.9% of enterprise tokens in November, up from 7.3% in July Databricks AI Index. Databricks notes a commercial interest in open-weight models and cautions that its customers may lean further toward them than the average enterprise (conflict of interest, own or unaudited numbers).
Enterprises relying on a single model provider dropped from roughly 64% in July to 28% by November, while the average number of unique models per customer rose from 4.2 to 7.1 since January.
Anthropic retained the lead among proprietary models, increasing its share from 57.6% to 68.8%. Its Opus 5.5 captured 37.9% of proprietary tokens within six weeks of launch, while Claude Fable 5.1 fell from 5.5% to 0.6% in the same period, underscoring price-performance trade-offs over raw capability.
Spending distribution is highly skewed: the median enterprise user spends $2.49 per week, more than doubling from $1.10 six weeks earlier, whereas the 99th-percentile user spends $699.74 weekly—a 281-fold difference.
Our read: The rapid rise of open-weight models and the collapse of single-provider dominance suggest enterprises are optimizing for cost and workload-specific performance, while the stark spend variance signals that per-seat pricing will increasingly misalign with actual usage.
Interested in implementing this technology inside your company?
Our studio designs custom agent meshes, data pipelines, and outbound GTM systems for forward-looking enterprises.