What's happened
Open-source AI models are gaining ground against closed systems. Z.ai's GLM 5.2 and DeepSeek V4 show strong capability at lower cost, while U.S. policy and international rivalry push firms toward self-hosted solutions. CNBC notes Anthropic’s model outages underscore the fragility of centralised access. The trend could reshape vendor choice and pricing.
What's behind the headline?
Context and trajectory
- Open-source models are closing the gap with frontier closed models, driven by cost and flexibility.
- China-led open-source ecosystems (DeepSeek, GLM, MiniMax) are increasingly adopted in both enterprise and developer stacks, affecting vendor lock-in dynamics.
- US firms warn that reliance on external access to top models could risk IP and control; this is prompting consideration of self-hosted deployments and internal data governance.
Implications for readers
- Expect more CIOs to evaluate open-source options for routine workloads, reserving premium performance for specialized tasks.
- Regulators may tighten export controls and data-sharing rules, influencing deployment choices.
What to watch
- The pace at which open-source models catch up in accuracy and safety features.
- The willingness of vendors to support self-hosted deployments at scale.
How we got here
Open-source models from China dominate in cost-effective AI tooling, with firms increasingly migrating workloads from pricier closed models to open options hosted on US or international infrastructure. The competition between US and Chinese open-source ecosystems accelerates as regulators and chip controls influence model deployment and access.
Our analysis
Axios notes that open-source Chinese models are good enough for most work at a fraction of the price, with migration taking months; Business Insider UK highlights GLM 5.2 as a potential daily-driver for coding and agentic workflows; CNBC discusses the risk of model access being cut and the push for self-hosted solutions.
Go deeper
- Will firms accelerate self-hosted AI deployments to reduce vendor risk?
- How will export controls influence the availability of open-source models in different regions?
- Which sectors are most at risk of disruption from the shift to open-source AI?
More on these topics
-
Anthropic - Artificial intelligence company
Anthropic PBC is a U.S.-based artificial intelligence startup public-benefit company, founded in 2021. It researches and develops AI to "study their safety properties at the technological frontier" and use this research to deploy safe, reliable models for