CICC has released a research report indicating that the intelligence level of open-source models has accelerated this year, with capabilities approaching those of top-tier closed-source models. As a result, market attention has expanded from hardware and closed-source models to include open-source developers, cloud vendors, and application providers. The firm holds a positive outlook on open-source and closed-source models jointly driving the realization of AGI and recommends domestic open-source model developers. It also favors cloud vendors, expecting a value re-rating driven by open-source model adoption, and recommends application providers, citing accelerated growth in FDE and AI applications fueled by open-source model penetration.

Open-source model token call volumes are trending upward, unlocking vast market potential

Open-source models, leveraging high cost-effectiveness and strong intelligence, have gained widespread recognition from users both domestically and internationally, leading to a steady rise in call volumes. CICC estimates that overseas revenue from open-source models will reach tens of billions of dollars this year, and by 2027, global open-source model call volumes could account for 50%–70% of total token usage, with their value representing 12%–30% of the large model market. The firm believes that the accelerated penetration of open-source models may drive down the unit price of tokens, but their open and accessible nature will unlock more use cases, potentially boosting overall token consumption significantly.

Open strategies enhance intelligence and accelerate commercialization

Open-source models adhere to the first principles of model development, upholding the Scaling Law while intensifying post-training and reinforcement learning to continuously improve intelligence. Notably, Kimi K3 and Qwen 3.8 have both reached parameter scales approaching 3 trillion. The openness of these models enables deployment across more high-value application scenarios, while rising intelligence levels grant developers greater pricing power, such as through price increases or revenue-sharing discussions with platform operators, thereby accelerating commercialization.

Open-source models foster ecosystem growth, boosting demand for cloud vendors, FDE, and application providers

The open-source model ecosystem involves numerous participants, including compute hardware providers, cloud vendors, inference and tuning platforms, and aggregation or application platforms. Technological advancements in open-source models will also drive commercialization for these stakeholders. Cloud vendors, for instance, see increased value from deploying open-source models, while FDE plays a crucial role in bridging model deployment with enterprise workflow understanding. Application providers, meanwhile, leverage open-source models to enhance innovation capabilities and penetration efficiency. As the application-side total addressable market expands, it will, in turn, feed back into the growth of both open-source and closed-source model value chains.

Risks

Potential risks include slower-than-expected AI technological progress, weaker-than-anticipated downstream demand, and intensifying competition among vendors.