The state of open source AI

Open source AI models are rapidly gaining ground against their closed-source competitors, with usage patterns shifting dramatically in favor of freely available alternatives. The [State of Open Source AI](https://stateofopensource.ai/) report reveals how community-driven development is reshaping the artificial intelligence industry and challenging the dominance of proprietary models from major tech companies.

The Numbers Tell the Story

Open source AI models have experienced explosive growth in recent months. On OpenRouter, a platform that tracks AI model usage, open models now command 63% market share compared to 37% for closed models - a complete reversal from just four months ago when closed models held a 60% advantage.

The scale of this shift becomes clearer when examining token processing volumes. Open models processed 888 billion tokens in aggregate on March 19th. By late July, that number had jumped to 4.19 trillion tokens - nearly a five-fold increase in just four months.

This surge reflects more than simple preference changes. Open source models offer users freedom from licensing fees and vendor lock-in, making them attractive to hyperscale cloud providers who can deploy them without ongoing payments to model creators.

Corporate Strategies Diverge

Major technology companies are taking different approaches to this open source wave. Apple has shown particular interest in smaller, efficient models that can run directly on devices rather than requiring cloud connectivity. This strategy aligns perfectly with open source availability since the company can modify and optimize models for their specific hardware without negotiating complex licensing agreements.

Meanwhile, companies like Meta have released their Llama model series under open licenses, betting that widespread adoption will benefit their broader ecosystem even without direct model monetization.

The hyperscaler cloud providers - Amazon, Google, and Microsoft - find themselves in an interesting position. They can offer open source models to customers without paying licensing fees to AI startups, potentially improving their margins while still charging for compute infrastructure.

Sustainability Questions Emerge

The report identifies significant challenges facing the open source AI community. Governance structures remain underdeveloped compared to traditional open source software projects. Many open source AI initiatives lack clear decision-making processes or sustainable funding models.

Training costs for frontier models continue to climb, creating barriers for community-led development of the most advanced systems. While companies can release pre-trained models as open source, the computational resources required to create them from scratch remain concentrated among well-funded organizations.

The community also grapples with questions about model safety, responsible release practices, and how to handle potentially harmful capabilities in openly available systems.

This shift toward open source AI puts direct pressure on companies like OpenAI and Anthropic whose business models depend on maintaining technological advantages through proprietary development, while making advanced AI capabilities accessible to organizations that previously couldn't afford them.