Artificial Intelligence: Citi’s Inference Ahead – Open-Source Moment

Research evidence excerpt

Artificial Intelligence: Citi’s Inference Ahead – Open-Source Moment

Viewpoint |

26 Jun 2026 06:07:01 ET │ 23 pages

Artificial Intelligence

Citi’s Inference Ahead – Open-Source Moment

CITI'S TAKE

Heath Terry AC

As regulators block access to some frontier models and slow the release of +1-212-723-4624

others, open models are gaining momentum as inference needs continue to heath.terry@citi.com

outstrip supply, driving enterprises to seek out alternatives to frontier

models for specific use-cases. While benchmarks are limited for measuring Shelby Spencer

value to the enterprise, our dashboard (Figure 1) shows the intelligence gap +1-212-816-0416

between proprietary and open-weight models has compressed in the past shelby.spencer@citi.com

2 weeks with the release of Z.ai’s GLM-5.2 model. The shift is fueling

investment in startups providing cost-efficient, open architectures for both Ashley Kim

software and hardware. While infrastructure deals continue to cement the +1-212-816-6689

industry’s long-term direction, talent constraints and data center ashley.kim@citi.com

construction challenges will impact the pace of development, favoring AI

talent-dense and compute-dense leaders, even as we expect the pricing Janna Withrow

beneficiaries of these bottlenecks to continue to outperform. +1-212-723-0439

janna.withrow@citi.com

Open model demand - The demand for open-weight and open-source models is

seeing a step-function increase as enterprises first reacted to rising costs and

second to increasing restrictions on model access. Open model developers, API

routers, and inference engines are experiencing surging demand, with Cohere

tripling its internal ARR projection for 2027, OpenRouter’s share of open-source