OpenFold's strategy isn't just to provide a free tool. By releasing its training code and data, it enables companies to create specialized versions by privately fine-tuning the model on their own proprietary data. This allows firms to maintain a competitive edge while leveraging a shared, open foundation.

Companies like Z.ai are not abandoning open source but using it strategically. They release lightweight models to attract developers and build a user base, while reserving their most powerful, agentic systems for proprietary, revenue-generating enterprise products, creating a clear monetization funnel.

For a hardware-centric company, open-sourcing its LLM is a strategic move. It serves as a powerful talent magnet for top AI engineers and invites a global community of developers to help integrate the model across Xiaomi's vast ecosystem of devices, accelerating innovation at low cost.

Nvidia is heavily investing in its own open-source models like Nemo Tron. This strategy ensures that as the open-source ecosystem grows, demand for its hardware also grows, positioning Nvidia's chips as the default platform and reducing reliance on closed-source model providers who act as intermediaries.

AI21 exemplifies a winning AI business model: give away the foundational model (Jamba) to drive adoption, then monetize a proprietary orchestration layer (Maestro) that helps enterprises manage multiple models for cost and performance, capturing value higher up the stack.

Z.AI and other Chinese labs recognize Western enterprises won't use their APIs due to trust and data concerns. By open-sourcing models, they bypass this barrier to gain developer adoption, global mindshare, and brand credibility, viewing it as a pragmatic go-to-market tactic rather than an ideological stance.

Contrary to past momentum, the most advanced AI startups are increasingly adopting and fine-tuning open-source models. This shift is driven by the need for cost-effective speed and deep customization as their workloads mature and scale.

Instead of just releasing model weights, NVIDIA is publishing 10 trillion tokens of training data, 15 reinforcement learning environments, and full evaluation recipes. This strategy empowers researchers and developers to fully reproduce, adapt, and build on their work, fostering a deep ecosystem around their hybrid architecture.

VLLM thrives by creating a multi-sided ecosystem where stakeholders contribute for their own self-interest. Model providers contribute to ensure their models run well. Silicon providers (NVIDIA, AMD) contribute to support their hardware. This flywheel effect establishes the platform as a de facto standard, benefiting the entire ecosystem.