Open-source models like Minimax 3 Pro and GLM 5.3 threaten Anthropic and OpenAI's dominance
2.7T parameter open-source models are accelerating—and enterprises are losing trust in proprietary AI.
The AI landscape is witnessing a 'Kimi moment' as open-source models from Chinese labs—Minimax 3 Pro (with 2.7 trillion parameters) and GLM 5.3—gain rapid traction. These models collectively accelerate progress, challenging the narrative that proprietary camps hold an unassailable lead. The sheer scale and openness give developers and enterprises transparency and flexibility that walled-garden providers cannot match.
Meanwhile, trust in Anthropic and OpenAI is eroding. The argument goes: to re-establish their moat and justify sky-high valuations, these companies must vertically integrate into applications (finance, ads, etc.) and potentially 'distill' client data. Enterprises, wary of handing over proprietary knowledge, are increasingly exploring open-source alternatives. This shift could reshape competitive dynamics, making open models the default choice for sensitive workloads.
- Minimax 3 Pro boasts 2.7T parameters, one of the largest open-source models released.
- Open-source progress is accelerating collectively, outpacing individual proprietary efforts.
- Enterprise trust in Anthropic/OpenAI drops as concerns grow over data extraction and vertical expansion.
Why It Matters
This signals a major trust shift—enterprises may favor open models to avoid data extraction by proprietary AI vendors.