• Anthropic’s Claude Opus 5 ranks as the most intelligent AI model, according to Bank of America’s new Frontier AI Tracker.
  • AI token prices dropped 9% month-over-month, driven by OpenAI price cuts, while GPU rental costs remain stable.
  • DeepSeek leads usage share at 30%, but Anthropic dominates AI spending at 65%.

The New Frontier AI Tracker

Bank of America has launched its Frontier AI Tracker, a comprehensive tool that monitors model intelligence, usage, token prices, and hardware costs. The inaugural data, released this week, provides a real-time snapshot of the competitive landscape.

Intelligence Rankings: Claude Takes the Crown

According to the tracker, Anthropic’s Claude Opus 5 tops the intelligence benchmarks, with Claude Fable 5 and OpenAI’s GPT-5.6 Sol close behind. Notably, Opus 5 is touted as cheaper per unit of intelligence, a critical factor for enterprises scaling AI deployments.

“We are seeing a shift toward efficiency,” said a Bank of America analyst on condition of anonymity. “The intelligence-to-cost ratio is becoming a key differentiator.”

Spending and Usage: A Tale of Two Leaders

Anthropic dominates AI spending, capturing a staggering 65% of total expenditure in the sector. In contrast, DeepSeek leads in usage share at 30%, highlighting a divergence between where enterprises allocate budgets and where actual workloads run. This suggests that while DeepSeek is popular for certain tasks, Anthropic’s models are preferred for high-stakes, sophisticated applications.

Cost Dynamics: Falling Token Prices, Stable Hardware

AI token prices fell 9% month-over-month, catalyzed by significant price cuts from OpenAI. These reductions are part of a broader competitive strategy to maintain market share amidst rising challengers. Meanwhile, GPU rental costs have remained broadly stable, offering some relief to startups and researchers who rely on cloud infrastructure.

“The price war is benefiting consumers,” noted a tech industry analyst. “But the stability in GPU costs indicates that hardware supply constraints are easing.”

Implications for Enterprises

The tracker’s launch comes at a critical juncture as enterprises increasingly evaluate AI vendors based on both performance and total cost of ownership. With intelligence benchmarks tightening and prices falling, procurement decisions are becoming more data-driven.

One enterprise IT manager, who asked not to be named, said, “We’ve been reassessing our AI stack. The new data helps us justify switching to more cost-effective models without compromising on quality.”

Looking Ahead

Bank of America’s Frontier AI Tracker is expected to be updated monthly, providing ongoing visibility into this fast-evolving market. As competition intensifies, we can anticipate further price adjustments and innovations aimed at pushing the intelligence-per-dollar envelope.

Correction: An earlier version of this article incorrectly stated that OpenAI’s GPT-5.6 Sol was ranked second. It is actually third, behind Claude Fable 5. The rankings have been updated to reflect the latest data.